{"id":29087,"date":"2026-03-22T14:38:36","date_gmt":"2026-03-22T06:38:36","guid":{"rendered":"https:\/\/aif.amtbbs.org\/?p=29087"},"modified":"2026-03-22T15:05:32","modified_gmt":"2026-03-22T07:05:32","slug":"%e7%8b%a0%e4%ba%ba%e6%8f%ad%e7%a7%98claudecode%e3%80%81cursor%e3%80%81openai%e6%99%ba%e8%83%bd%e4%bd%93%e5%b7%a5%e7%a8%8b%e6%8a%80%e6%9c%af%ef%bc%9a%e6%a8%a1%e5%9e%8b%e5%87%a0%e4%b9%8e%e6%97%a0","status":"publish","type":"post","link":"https:\/\/aif.amtbbs.org\/index.php\/2026\/03\/22\/29087\/","title":{"rendered":"\u72e0\u4eba\u63ed\u79d8ClaudeCode\u3001Cursor\u3001OpenAI\u667a\u80fd\u4f53\u5de5\u7a0b\u6280\u672f\uff1a\u6a21\u578b\u51e0\u4e4e\u65e0\u5173\u7d27\u8981\uff0c\u800c\u662f\u6784\u5efa\u6b63\u786e\u7684\u5de5\u7a0b\u73af\u5883\uff01Harness\u5c31\u662f\u4e00\u5207\uff01"},"content":{"rendered":"<div class=\"article-desc\">\u53d8\u9769\u6027\u6280\u672f\u5728\u65e9\u671f\u9636\u6bb5\u5982\u4f55\u88ab\u8bef\u89e3\u662f\u6709\u6a21\u5f0f\u53ef\u5faa\u7684\u3002\u90a3\u4e9b\u5438\u5f15\u516c\u4f17\u6ce8\u610f\u529b\u7684\u4e1c\u897f\u2014\u2014\u539f\u751f\u80fd\u529b\u3001\u4ee4\u4eba\u5370\u8c61\u6df1\u523b\u7684\u6f14\u793a\u3001\u57fa\u51c6\u6d4b\u8bd5\u5206\u6570\u2014\u2014\u5f88\u5c11\u662f\u51b3\u5b9a\u957f\u671f\u80dc\u8d1f\u7684\u5173\u952e\u3002\u57fa\u7840\u8bbe\u65bd\u5c42\u3001Harness\u3001\u73af\u5883\uff0c\u901a\u5e38\u624d\u662f\u771f\u6b63\u4ef7\u503c\u4ea7\u751f\u548c\u88ab\u6355\u83b7\u7684\u5730\u65b9\u3002<\/div>\n<div id=\"postspictures\" class=\"article-content\">\n<div id=\"container\" class=\"container am-engine\" data-v-01a18e2f=\"\" data-element=\"root\">\n<p>\u7f16\u8f91 | \u7389\u6f84<\/p>\n<p>\u51fa\u54c1 | 51CTO\u6280\u672f\u6808\uff08\u5fae\u4fe1\u53f7\uff1ablog51cto\uff09<\/p>\n<p>2026\u5e74\u5f00\u5e74\u4ee5\u6765\uff0cHarness\u5de5\u7a0b\u4e00\u8bcd\u70ed\u5ea6\u6e10\u9ad8\uff0cOpenAI\u57282\u6708\u53d1\u5e03\u7684\u4e00\u7bc7\u8be6\u7ec6\u7684\u5185\u90e8\u5b9e\u9a8c\u62a5\u544a\u6807\u9898\u4e2d\u4f7f\u7528\u4e86\u6b64\u8bcd\uff0cThoughtWorks \u9996\u5e2d\u79d1\u5b66\u5bb6\u00a0Martin Fowler \u5728 X\u4e0a\u4e5f\u8868\u793aHarness\u5de5\u7a0b\u662fAI\u8d4b\u80fd\u8f6f\u4ef6\u5f00\u53d1\u7684\u5173\u952e\u90e8\u5206\u3002<\/p>\n<p><img data-dominant-color=\"dbdce0\" data-has-transparency=\"true\" style=\"--dominant-color: #dbdce0;\" loading=\"lazy\" decoding=\"async\" class=\"has-transparency alignnone size-full wp-image-29097\" src=\"https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/a38e0b594ea8f1c4f34507b8615a1445b0b06a-1.png\" width=\"875\" height=\"447\" alt=\"\u56fe\u7247\" title=\"\u56fe\u7247\" srcset=\"https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/a38e0b594ea8f1c4f34507b8615a1445b0b06a-1.png 875w, https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/a38e0b594ea8f1c4f34507b8615a1445b0b06a-1-300x153.png 300w, https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/a38e0b594ea8f1c4f34507b8615a1445b0b06a-1-150x77.png 150w, https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/a38e0b594ea8f1c4f34507b8615a1445b0b06a-1-768x392.png 768w\" sizes=\"auto, (max-width: 875px) 100vw, 875px\" \/><span class=\"data-image-title\">\u56fe\u7247<\/span><\/p>\n<p>OpenAI\u6587\u7ae0\u94fe\u63a5\uff1ahttps:\/\/openai.com\/index\/harness-engineering\/<\/p>\n<p>\u8fd1\u65e5\uff0c\u4e00\u4f4d\u8d44\u6df1\u5168\u6808\u5de5\u7a0b\u5e08\u5728X\u4e0a\u53d1\u5e03\u4e86\u4e00\u7bc7\u540d\u4e3a\u300aHarness \u5c31\u662f\u4e00\u5207\uff1aCursor\u3001Claude Code \u548c Perplexity \u5230\u5e95\u6784\u5efa\u4e86\u4ec0\u4e48\u300b\u7684\u6587\u7ae0\u3002\u4ed6\u5c06Harness\u4e4b\u4e8eAgent\uff0c\u7c7b\u6bd4\u4e8e\u5e94\u7528\u5546\u5e97\u548c\u5f00\u53d1\u5de5\u5177\u4e4b\u4e8e\u79fb\u52a8\u7aef\u3001\u641c\u7d22\u5f15\u64ce\u548c\u6d4f\u89c8\u5668\u4e4b\u4e8e\u4e92\u8054\u7f51\uff0c\u8ba4\u4e3a\u662f\u76ee\u524d\u5e94\u7528AI\u9886\u57df\u6700\u6838\u5fc3\u7684\u5de5\u7a0b\u95ee\u9898\u3002<\/p>\n<p><img data-dominant-color=\"8d9496\" data-has-transparency=\"true\" style=\"--dominant-color: #8d9496;\" loading=\"lazy\" decoding=\"async\" class=\"has-transparency alignnone size-full wp-image-29098\" src=\"https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/87d343c9060eb37c63a434b142009e6ec91d6d-1.png\" width=\"889\" height=\"647\" alt=\"\u56fe\u7247\" title=\"\u56fe\u7247\" srcset=\"https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/87d343c9060eb37c63a434b142009e6ec91d6d-1.png 889w, https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/87d343c9060eb37c63a434b142009e6ec91d6d-1-300x218.png 300w, https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/87d343c9060eb37c63a434b142009e6ec91d6d-1-150x109.png 150w, https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/87d343c9060eb37c63a434b142009e6ec91d6d-1-768x559.png 768w\" sizes=\"auto, (max-width: 889px) 100vw, 889px\" \/><span class=\"data-image-title\">\u56fe\u7247<\/span><\/p>\n<p>\u6709\u4eba\u5728\u6587\u7ae0\u4e0b\u65b9\u8bc4\u8bba\uff1a\u201c\u8fd9\u662fAI\u9886\u57df\u7684\u4e0b\u4e00\u4e2a\u5343\u4ebf\u7ea7\u7684\u673a\u4f1a\uff0c\u5927\u591a\u6570\u4eba\u4f1a\u9519\u5931\uff0c\u5b83\u5c31\u662fHarness\u5de5\u7a0b\u3002\u201d<\/p>\n<p><img data-dominant-color=\"eceded\" data-has-transparency=\"false\" style=\"--dominant-color: #eceded;\" loading=\"lazy\" decoding=\"async\" class=\"not-transparent alignnone size-full wp-image-29099\" src=\"https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/b4688c921db0e0e129b47709fbe9d643effb67-1.jpg\" width=\"881\" height=\"537\" alt=\"\u56fe\u7247\" title=\"\u56fe\u7247\" srcset=\"https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/b4688c921db0e0e129b47709fbe9d643effb67-1.jpg 881w, https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/b4688c921db0e0e129b47709fbe9d643effb67-1-300x183.jpg 300w, https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/b4688c921db0e0e129b47709fbe9d643effb67-1-150x91.jpg 150w, https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/b4688c921db0e0e129b47709fbe9d643effb67-1-768x468.jpg 768w\" sizes=\"auto, (max-width: 881px) 100vw, 881px\" \/><span class=\"data-image-title\">\u56fe\u7247<\/span><\/p>\n<p>\u65e0\u72ec\u6709\u5076\uff0c\u57283\u670818\u53f7\uff0cMiniMax\u53d1\u5e03\u81ea\u5bb6\u7b2c\u4e00\u4e2a\u6a21\u578b\u6df1\u5ea6\u53c2\u4e0e\u8fed\u4ee3\u81ea\u5df1\u7684M2.7\u6a21\u578b\u65f6\uff0c\u7740\u91cd\u5f3a\u8c03\u4e86M2.7\u80fd\u591f\u81ea\u4e3b\u6784\u5efaAgent Harness\uff0c\u5b8c\u6210\u9ad8\u5ea6\u590d\u6742\u7684\u751f\u4ea7\u529b\u4efb\u52a1\u3002<\/p>\n<p><img data-dominant-color=\"eaf2ec\" data-has-transparency=\"false\" style=\"--dominant-color: #eaf2ec;\" loading=\"lazy\" decoding=\"async\" class=\"not-transparent alignnone size-full wp-image-29100\" src=\"https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/13d683258b7bd4f2c7a9459fa0922675c7024b-1.png\" width=\"1080\" height=\"568\" alt=\"\u56fe\u7247\" title=\"\u56fe\u7247\" srcset=\"https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/13d683258b7bd4f2c7a9459fa0922675c7024b-1.png 1080w, https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/13d683258b7bd4f2c7a9459fa0922675c7024b-1-300x158.png 300w, https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/13d683258b7bd4f2c7a9459fa0922675c7024b-1-1024x539.png 1024w, https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/13d683258b7bd4f2c7a9459fa0922675c7024b-1-150x79.png 150w, https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2026\/03\/13d683258b7bd4f2c7a9459fa0922675c7024b-1-768x404.png 768w\" sizes=\"auto, (max-width: 1080px) 100vw, 1080px\" \/><span class=\"data-image-title\">\u56fe\u7247<\/span><\/p>\n<p>\u4ee5\u4e0b\u4e3a\u4e07\u5b57\u539f\u6587\u5bf9Harness\u7684\u8be6\u7ec6\u9610\u8ff0\uff1a<\/p>\n<h3>AI\u7528\u5f97\u4e0d\u597d\uff0c\u662f\u56e0\u4e3a\u73af\u5883\u6ca1\u6784\u5efa\u5bf9<\/h3>\n<p>\u4f60\u4e4b\u6240\u4ee5\u89c9\u5f97 AI \u7528\u5f97\u4e0d\u5bf9\uff0c\u5e76\u4e0d\u662f\u56e0\u4e3a\u8fd8\u6ca1\u627e\u5230\u6b63\u786e\u7684\u6a21\u578b\u3002\u4f60\u7528\u9519 AI \u7684\u539f\u56e0\u662f\uff0c\u4f60\u6ca1\u6709\u6784\u5efa\u6b63\u786e\u7684<strong>\u73af\u5883<\/strong>\u3002<\/p>\n<p>\u8fd9\u5c31\u662f\u4e3a\u4ec0\u4e48\u6709\u4e9b\u56e2\u961f\u4ec5\u51ed\u4e09\u540d\u5de5\u7a0b\u5e08\u5c31\u80fd\u4ea4\u4ed8\u6570\u767e\u4e07\u884c\u4ee3\u7801\uff0c\u800c\u5176\u4ed6\u56e2\u961f\u751a\u81f3\u8fde\u8ba9\u667a\u80fd\u4f53\u6d41\u6c34\u7ebf\u5b8c\u6210\u4e00\u6b21\u8fde\u8d2f\u7684\u91cd\u6784\u90fd\u56f0\u96be\u91cd\u91cd\u3002<\/p>\n<p>\u8fd9\u79cd\u5dee\u8ddd\u4e0d\u5728\u4e8e GPT-5 \u4e0e Claude Opus \u7684\u533a\u522b\uff0c\u4e5f\u4e0d\u5728\u4e8e\u6e29\u5ea6\u8bbe\u7f6e\u6216\u6700\u5927 Token \u6570 \u3002\u751a\u81f3\u4e0d\u5728\u4e8e\u63d0\u793a\u8bcd\uff0c\u5c3d\u7ba1\u6bcf\u4e2a\u4eba\u90fd\u4e3a\u4e86\u63d0\u793a\u8bcd\u4e89\u8bba\u4e0d\u4f11\uff0c\u6d6a\u8d39\u4e86\u6570\u6708\u7684\u4eba\u751f\u3002<\/p>\n<p><strong>\u771f\u6b63\u7684\u533a\u522b\u5728\u4e8e Harness<\/strong>\u3002<\/p>\n<p>\u8fd9\u7bc7\u6587\u7ae0\u5c06\u63a2\u8ba8\u8fd9\u4e2a\u8bcd\u5728\u6280\u672f\u548c\u54f2\u5b66\u4e0a\u7684\u771f\u6b63\u542b\u4e49\uff0c\u56e0\u4e3a\u4e1a\u754c\u5df2\u7ecf\u517b\u6210\u4e86\u4e00\u79cd\u968f\u610f\u4f7f\u7528\u5b83\u7684\u574f\u4e60\u60ef\u3002Harness \u4e0d\u662f\u7cfb\u7edf\u63d0\u793a\u8bcd\uff0c\u4e0d\u662f API \u8c03\u7528\u7684\u5305\u88c5\u5668 \u3002\u5b83\u4e0d\u662f\u8bc4\u6d4b\u6846\u67b6\u3001\u63d0\u793a\u8bcd\u6a21\u677f\uff0c\u4e5f\u4e0d\u662f\u5e26\u8bb0\u5fc6\u7684\u804a\u5929\u673a\u5668\u4eba\u3002<\/p>\n<p><strong>Harness \u662f\u8bed\u8a00\u6a21\u578b\u8fd0\u884c\u7684\u5b8c\u6574\u8bbe\u8ba1\u73af\u5883<\/strong>\uff0c\u5305\u62ec\uff1a\u5b83\u53ef\u4ee5\u8c03\u7528\u7684\u5de5\u5177\u3001\u5b83\u63a5\u6536\u4fe1\u606f\u7684\u683c\u5f0f\u3001\u5b83\u5386\u53f2\u8bb0\u5f55\u7684\u538b\u7f29\u4e0e\u7ba1\u7406\u65b9\u5f0f\u3001\u5728\u9519\u8bef\u7ea7\u8054\u524d\u62e6\u622a\u9519\u8bef\u7684\u62a4\u680f\uff0c\u4ee5\u53ca\u5141\u8bb8\u5b83\u5c06\u5de5\u4f5c\u79fb\u4ea4\u7ed9\u201c\u672a\u6765\u7684\u81ea\u5df1\u201d\u800c\u53c8\u4e0d\u4e22\u5931\u8fde\u8d2f\u6027\u7684\u811a\u624b\u67b6\u3002<\/p>\n<p>\u5f53\u4f60\u5ba1\u89c6 Anthropic \u4e3a\u4f7f Claude Code \u771f\u6b63\u843d\u5730\u800c\u6784\u5efa\u7684\u5185\u5bb9\u3001OpenAI \u5982\u4f55\u901a\u8fc7 Codex \u4ea4\u4ed8\u6570\u767e\u4e07\u884c\u96f6\u4eba\u5de5\u7f16\u5199\u7684\u4ee3\u7801\uff0c\u4ee5\u53ca\u666e\u6797\u65af\u987f NLP \u5c0f\u7ec4\u5728\u91cc\u7a0b\u7891\u5f0f\u7684 SWE-agent \u8bba\u6587\u4e2d\u53d1\u8868\u7684\u5173\u4e8e\u201c\u667a\u80fd\u4f53-\u8ba1\u7b97\u673a\u63a5\u53e3\uff08ACI\uff09\u201d\u7684\u7814\u7a76\u65f6\uff0c\u4f60\u4f1a\u53d1\u73b0\u6240\u6709\u4e25\u8083\u5bf9\u5f85\u8fd9\u4e00\u9886\u57df\u7684\u56e2\u961f\u90fd\u5728\u5448\u73b0\u540c\u4e00\u79cd\u6a21\u5f0f\u3002<\/p>\n<p><strong>\u6a21\u578b\u51e0\u4e4e\u65e0\u5173\u7d27\u8981\uff0cHarness \u5c31\u662f\u4e00\u5207<\/strong>\u3002<\/p>\n<p>\u8fd9\u662f\u5173\u4e8e\u8be5\u7406\u5ff5\u5982\u4f55\u6210\u4e3a 2025 \u548c 2026 \u5e74\u5e94\u7528 AI \u5de5\u7a0b\u5b9a\u4e49\u7c7b\u6d1e\u5bdf\u7684\u8be6\u7ec6\u6280\u672f\u5206\u89e3\u3002\u5b83\u6db5\u76d6\u4e86\u7814\u7a76\u3001\u5b9e\u9645\u5b9e\u73b0\u3001\u9a71\u52a8\u8bbe\u8ba1\u51b3\u7b56\u7684\u5931\u8d25\u6a21\u5f0f\uff0c\u4ee5\u53ca\u65e0\u8bba\u4f60\u662f\u5728\u6784\u5efa\u7f16\u7801\u667a\u80fd\u4f53\u3001\u7814\u7a76\u667a\u80fd\u4f53\u8fd8\u662f\u957f\u671f\u8fd0\u884c\u7684\u81ea\u4e3b\u8f6f\u4ef6\u5de5\u7a0b\u5e08\u65f6\u90fd\u4f1a\u91cd\u590d\u51fa\u73b0\u7684\u6a21\u5f0f \u3002<\/p>\n<p>\u8bfb\u5b8c\u672c\u6587\uff0c\u4f60\u4e0d\u4ec5\u4f1a\u7406\u89e3\u4ec0\u4e48\u662f Harness\uff0c\u8fd8\u4f1a\u660e\u767d\u4e3a\u4ec0\u4e48\u6b63\u786e\u6784\u5efa\u5b83\u5df2\u6210\u4e3a\u884c\u4e1a\u5185\u6700\u6709\u4ef7\u503c\u7684\u5de5\u7a0b\u6280\u80fd \u3002<\/p>\n<h3>\u6ca1\u4eba\u8ba8\u8bba\u7684\u95ee\u9898<\/h3>\n<h4>\u4e3a\u4ec0\u4e48\u539f\u59cb\u80fd\u529b\u662f\u4e0d\u591f\u7684<\/h4>\n<p>2024 \u5e74\u4e2d\u671f\uff0cAI \u57fa\u51c6\u6d4b\u8bd5\u4e2d\u53d1\u751f\u4e86\u4e00\u4ef6\u5947\u602a\u7684\u4e8b \u3002\u7814\u7a76\u4eba\u5458\u53d1\u73b0\uff0c\u540c\u4e00\u4e2a\u524d\u6cbf\u6a21\u578b\u5728\u76f8\u540c\u7684\u7f16\u7801\u4efb\u52a1\u4e0a\uff0c\u7531\u4e8e\u4efb\u52a1\u5448\u73b0\u65b9\u5f0f\u548c\u53ef\u7528\u5de5\u5177\u7684\u4e0d\u540c\uff0c\u4f1a\u4ea7\u751f\u622a\u7136\u4e0d\u540c\u7684\u7ed3\u679c\u3002\u6a21\u578b\u6ca1\u53d8\uff0c\u5e95\u5c42\u7684\u667a\u80fd\u6ca1\u53d8\uff0c\u6539\u53d8\u7684\u662f<strong>\u63a5\u53e3<\/strong>\u3002<\/p>\n<p>\u8fd9\u672c\u4e0d\u8be5\u8ba9\u4eba\u611f\u5230\u60ca\u8bb6\u3002\u51e0\u5341\u5e74\u6765\u6211\u4eec\u90fd\u77e5\u9053\uff0c\u5408\u9002\u7684\u5de5\u5177\u80fd\u8ba9\u5de5\u7a0b\u5e08\u7684\u751f\u4ea7\u529b\u4ea7\u751f\u8d28\u7684\u98de\u8dc3\u3002\u4e00\u4e2a\u62e5\u6709\u73b0\u4ee3 IDE\u3001\u8c03\u8bd5\u5668\u3001\u7248\u672c\u63a7\u5236\u548c CI\/CD \u6d41\u6c34\u7ebf\u7684\u8f6f\u4ef6\u5f00\u53d1\u4eba\u5458\uff0c\u5176\u6548\u7387\u6bd4\u4ec5\u5728\u539f\u59cb\u7ec8\u7aef\u4f7f\u7528\u6587\u672c\u7f16\u8f91\u5668\u7684\u540c\u4e00\u4eba\u9ad8\u51fa\u51e0\u4e2a\u6570\u91cf\u7ea7\u3002IDE \u5e76\u6ca1\u6709\u8ba9\u5f00\u53d1\u4eba\u5458\u53d8\u806a\u660e\uff0c\u5b83\u53ea\u662f\u51cf\u5c11\u4e86\u6469\u64e6\u3001\u5728\u6b63\u786e\u65f6\u523b\u5448\u73b0\u4fe1\u606f\u3001\u53ca\u65e9\u53d1\u73b0\u9519\u8bef\u5e76\u7ec4\u7ec7\u5de5\u4f5c\u5355\u5143\u3002<\/p>\n<p>\u8bed\u8a00\u6a21\u578b\u4e5f\u662f\u5982\u6b64 \u3002\u5b83\u4eec\u5e76\u4e0d\u662f\u57fa\u4e8e\u65e0\u9650\u5185\u90e8\u77e5\u8bc6\u5e93\u5de5\u4f5c\u7684\u901a\u7528\u63a8\u7406\u8005\u3002\u5b83\u4eec\u662f\u8fd0\u4f5c\u5728\u4e0a\u4e0b\u6587\u7a97\u53e3\u4e2d\u7684\u7cbe\u5bc6<strong>\u6a21\u5f0f\u5339\u914d\u5f15\u64ce<\/strong>\u3002\u5b83\u4eec\u5728\u7279\u5b9a\u65f6\u523b\u6240\u77e5\u9053\u7684\u4e00\u5207\uff0c\u90fd\u53d6\u51b3\u4e8e\u7a97\u53e3\u4e2d\u7684\u5185\u5bb9\uff1b\u5b83\u4eec\u4ea7\u51fa\u7684\u4e00\u5207\uff0c\u90fd\u53d7\u9650\u4e8e\u8fd9\u4e9b\u4e0a\u4e0b\u6587\u7684\u7ed3\u6784\u3002\u8f93\u5165\u683c\u5f0f\u4e0d\u662f\u88c5\u9970\uff0c\u5b83\u662f\u667a\u80fd\u4f53\u7684<strong>\u8ba4\u77e5\u67b6\u6784<\/strong>\u3002\u63a5\u53e3\u4e0d\u662f\u4fbf\u5229\u5c42\uff1b\u5bf9\u4e8e\u8bed\u8a00\u6a21\u578b\u667a\u80fd\u4f53\u6765\u8bf4\uff0c<strong>\u63a5\u53e3\u5373\u601d\u60f3<\/strong>\u3002<\/p>\n<p>\u8fd9\u662f\u666e\u6797\u65af\u987f NLP \u5c0f\u7ec4 2024 \u5e74\u53d1\u8868\u7684 SWE-agent \u8bba\u6587\u7684\u6838\u5fc3\u4e3b\u5f20\uff0c\u4e14\u7ecf\u53d7\u4f4f\u4e86\u63a8\u6572\u3002\u8be5\u8bba\u6587\u5f15\u5165\u4e86<strong>\u667a\u80fd\u4f53-\u8ba1\u7b97\u673a\u63a5\u53e3 (Agent-Computer Interface, ACI)<\/strong>\u00a0\u7684\u6982\u5ff5\uff0c\u5e76\u8bc1\u660e\u4e86\u7cbe\u5fc3\u8bbe\u8ba1\u7684 ACI \u4e0e\u6807\u51c6 Linux Shell \u76f8\u6bd4\uff0c\u80fd\u8ba9\u540c\u4e00\u6a21\u578b\u5728\u57fa\u51c6\u6d4b\u8bd5\u4e2d\u7684\u6027\u80fd\u4ea7\u751f\u00a0<strong>64%<\/strong>\u00a0\u7684\u76f8\u5bf9\u63d0\u5347\u3002\u76f8\u540c\u7684\u6a21\u578b\uff0c\u76f8\u540c\u7684\u4efb\u52a1\uff0c\u76f8\u540c\u7684\u8ba1\u7b97\u9884\u7b97\uff0c\u53d8\u91cf\u53ea\u6709\u63a5\u53e3\u3002<\/p>\n<p>64% \u4e0d\u662f\u8fb9\u9645\u6536\u76ca\uff0c\u800c\u662f\u5de5\u5177\u201c\u80fd\u7528\u201d\u4e0e\u201c\u4e0d\u80fd\u7528\u201d\u7684\u533a\u522b\u3002\u800c\u8fd9\u5b8c\u5168\u6e90\u4e8e\u73af\u5883\u8bbe\u8ba1\uff0c\u800c\u975e\u5e95\u5c42\u6a21\u578b\u7684\u6539\u8fdb\u3002<\/p>\n<h4>\u4e0a\u4e0b\u6587\u7a97\u53e3\u4e0d\u662f\u5185\u5b58\u6761<\/h4>\n<p>\u5173\u4e8e AI \u667a\u80fd\u4f53\u7684\u4e00\u79cd\u539f\u751f\u6a21\u578b\u662f\u5c06\u4e0a\u4e0b\u6587\u7a97\u53e3\u89c6\u4e3a RAM\uff08\u968f\u673a\u5b58\u53d6\u5b58\u50a8\u5668\uff09\u3002\u4f60\u52a0\u8f7d\u6570\u636e\uff0c\u6a21\u578b\u5904\u7406\uff0c\u7136\u540e\u8f93\u51fa\u3002\u8fd9\u79cd\u89c2\u70b9\u8ba4\u4e3a\uff1a\u66f4\u591a\u7684\u4e0a\u4e0b\u6587\u7b49\u4e8e\u66f4\u597d\u7684\u6027\u80fd\uff0c\u66f4\u957f\u7684\u63d0\u793a\u8bcd\u7b49\u4e8e\u66f4\u6df1\u7684\u7406\u89e3 \u3002\u8fd9\u79cd\u89c2\u70b9\u662f\u9519\u8bef\u7684\uff0c\u5982\u679c\u4f60\u56f4\u7ed5\u5b83\u6784\u5efa\u667a\u80fd\u4f53\uff0c\u5b83\u4f1a\u6bc1\u4e86\u4f60\u7684\u5de5\u4f5c \u3002<\/p>\n<p>\u4e0a\u4e0b\u6587\u7a97\u53e3\u66f4\u50cf\u662f\u667a\u80fd\u4f53\u5728\u7ed9\u5b9a\u4f1a\u8bdd\u4e2d\u7684<strong>\u6574\u4e2a\u5de5\u4f5c\u610f\u8bc6<\/strong>\u3002\u7a97\u53e3\u4e2d\u7684\u6bcf\u4e2a Token \u90fd\u6709\u8ba1\u7b97\u6210\u672c\u3002\u6bcf\u4e00\u6761\u65e0\u5173\u4fe1\u606f\u90fd\u4f1a\u4e0e\u5173\u952e\u4fe1\u606f\u4e89\u593a\u6ce8\u610f\u529b\u3002\u6a21\u578b\u5e76\u6ca1\u6709\u4e00\u79cd\u80fd\u5e72\u51c0\u5229\u843d\u5730\u5ffd\u7565\u566a\u97f3\u7684\u9009\u62e9\u6027\u6ce8\u610f\u529b\u673a\u5236\u3002\u566a\u97f3\u5c31\u5728\u623f\u95f4\u91cc\uff0c\u5b83\u4f1a\u5f71\u54cd\u63a8\u7406\u3002<\/p>\n<p>\u8fd9\u5bf9\u667a\u80fd\u4f53\u8bbe\u8ba1\u6709\u5177\u4f53\u4e14\u53ef\u8861\u91cf\u7684\u540e\u679c\u3002\u5f53\u4f60\u4ece\u667a\u80fd\u4f53\u5faa\u73af\u5185\u90e8\u5728\u5927\u578b\u4ee3\u7801\u5e93\u8fd0\u884c grep \u5e76\u8fd4\u56de\u4e00\u4e07\u884c\u5339\u914d\u9879\u65f6\uff0c\u4f60\u5e76\u6ca1\u6709\u7ed9\u667a\u80fd\u4f53\u66f4\u591a\u4fe1\u606f\uff0c\u800c\u662f\u7528\u65e0\u5173\u6570\u636e\u6df9\u6ca1\u4e86\u5b83\u7684\u5de5\u4f5c\u8bb0\u5fc6\uff0c\u8fd9\u5c06\u964d\u4f4e\u5176\u540e\u6bcf\u4e00\u6b65\u7684\u8d28\u91cf\uff0c\u76f4\u5230\u4e0a\u4e0b\u6587\u88ab\u6e05\u9664\u3002\u5f53\u4f60\u56e0\u4e3a\u667a\u80fd\u4f53\u60f3\u770b\u4e24\u4e2a\u51fd\u6570\u5c31\u7528 cat \u503e\u5012\u6574\u4e2a\u6587\u4ef6\u65f6\uff0c\u4f60\u662f\u5728\u5b83\u9700\u8981\u4e00\u676f\u6c34\u7684\u65f6\u5019\u7ed9\u4e86\u5b83\u4e00\u652f\u9ad8\u538b\u6c34\u67aa\u3002<\/p>\n<p>SWE-agent \u7684\u7814\u7a76\u4eba\u5458\u8be6\u7ec6\u8bb0\u5f55\u4e86\u8fd9\u4e9b\u5931\u8d25\u6a21\u5f0f\u3002\u6807\u51c6 Bash \u63a5\u53e3\u4f1a\u5bfc\u81f4\u667a\u80fd\u4f53\u201c\u7a7a\u8f6c\u201d\uff1a\u5b83\u4eec\u4f1a\u53d1\u5e03\u8fd4\u56de\u6570\u5343\u884c\u7684 grep \u547d\u4ee4\uff0c\u5fd8\u8bb0\u81ea\u5df1\u8981\u627e\u4ec0\u4e48\uff0c\u63a5\u7740\u53d1\u5e03\u66f4\u591a grep\uff0c\u5bfc\u81f4\u4e0a\u4e0b\u6587\u5145\u6ee1\u566a\u97f3\uff0c\u6700\u7ec8\u4ea7\u751f\u9519\u8bef\u7b54\u6848\u6216\u505c\u6ede\u4e0d\u524d\u3002\u95ee\u9898\u4e0d\u5728\u4e8e\u6a21\u578b\u667a\u80fd\uff0c\u800c\u5728\u4e8e\u63a5\u53e3\u6ca1\u6709<strong>\u4fdd\u62a4\u667a\u80fd\u4f53\u514d\u53d7\u5176\u81ea\u8eab\u4f24\u5bb3<\/strong>\u7684\u673a\u5236 \u3002<\/p>\n<p>ACI \u7684\u89e3\u51b3\u65b9\u6848\u662f\u6784\u5efa\u4e00\u4e2a\u6709\u8fd4\u56de<strong>\u4e0a\u9650\u4e14\u80fd\u603b\u7ed3<\/strong>\u7684\u641c\u7d22\u5de5\u5177\u3002\u5982\u679c\u641c\u7d22\u8fd4\u56de\u8d85\u8fc7 50 \u4e2a\u5339\u914d\u9879\uff0c\u5de5\u5177\u4f1a\u6291\u5236\u8f93\u51fa\u5e76\u544a\u8bc9\u667a\u80fd\u4f53\u7f29\u5c0f\u67e5\u8be2\u8303\u56f4\u3002\u8fd9\u4e2a\u56de\u60f3\u8d77\u6765\u7b80\u5355\u5f97\u8fd1\u4e4e\u53d7\u8fb1\u7684\u8bbe\u8ba1\u51b3\u7b56\uff0c\u662f\u8bba\u6587\u4e2d\u6760\u6746\u7387\u6700\u9ad8\u7684\u53d8\u5316\u4e4b\u4e00\u3002\u5b83\u5c06\u201c\u4e0a\u4e0b\u6587\u6df9\u6ca1\u201d\u8fd9\u79cd\u5931\u8d25\u6a21\u5f0f\u8f6c\u5316\u4e3a\u4e86\u81ea\u7136\u7684\u6539\u8fdb\u5faa\u73af \u3002<\/p>\n<h3>SWE-Agent \u8bba\u6587\u4e0e\u00a0ACI\u00a0\u7684\u8bde\u751f<\/h3>\n<h4>\u667a\u80fd\u4f53-\u8ba1\u7b97\u673a\u63a5\u53e3\uff08ACI\uff09\u7a76\u7adf\u662f\u4ec0\u4e48<\/h4>\n<p>\u5728 SWE-agent \u8bba\u6587\u4e2d\uff0cACI \u88ab\u5b9a\u4e49\u4e3a\u4f4d\u4e8e\u8bed\u8a00\u6a21\u578b\u667a\u80fd\u4f53\u4e0e\u8ba1\u7b97\u673a\u73af\u5883\u4e4b\u95f4\u7684\u62bd\u8c61\u5c42\u3002\u5c06\u5176\u7c7b\u6bd4\u4e8e\u4eba\u673a\u4ea4\u4e92\uff08HCI\uff09\u662f\u523b\u610f\u4e3a\u4e4b\u7684\u3002\u6b63\u5982 HCI \u7814\u7a76\u5982\u4f55\u8bbe\u8ba1\u7b26\u5408\u4eba\u7c7b\u8ba4\u77e5\u67b6\u6784\u7684\u63a5\u53e3\u4e00\u6837\uff0cACI \u7814\u7a76\u5219\u63a2\u8ba8\u5982\u4f55\u8bbe\u8ba1\u7b26\u5408\u8bed\u8a00\u6a21\u578b\uff08LM\uff09\u8ba4\u77e5\u67b6\u6784\u7684\u63a5\u53e3\u3002<\/p>\n<p>\u4eba\u7c7b\u7684\u8ba4\u77e5\u67b6\u6784\u6d89\u53ca\u89c6\u89c9\u6a21\u5f0f\u8bc6\u522b\u3001\u7a7a\u95f4\u8bb0\u5fc6\u3001\u5c4f\u5e55\u4e0a\u7684\u5e76\u884c\u6ce8\u610f\uff0c\u4ee5\u53ca\u7565\u8bfb\u548c\u9009\u62e9\u6027\u5173\u6ce8\u7684\u80fd\u529b\u3002LM \u7684\u8ba4\u77e5\u67b6\u6784\u5219\u6839\u672c\u4e0d\u540c\uff1a\u5b83\u6d89\u53ca\u5e8f\u5217\u5316\u7684 Token \u5904\u7406\u3001\u5bf9\u4e0a\u4e0b\u6587\u987a\u5e8f\u548c\u683c\u5f0f\u7684\u654f\u611f\u6027\u3001\u6709\u9650\u7684\u5de5\u4f5c\u8bb0\u5fc6\uff0c\u4ee5\u53ca\u503e\u5411\u4e8e\u951a\u5b9a\u5728\u63d0\u793a\u8bcd\u4e2d\u6700\u663e\u8457\u4fe1\u606f\u7684\u503e\u5411\u3002\u8bbe\u8ba1\u4e00\u4e2a\u597d\u7684 ACI \u610f\u5473\u7740\u7406\u89e3\u8fd9\u4e9b\u7ea6\u675f\u5e76\u56f4\u7ed5\u5b83\u4eec\u8fdb\u884c\u6784\u5efa\uff0c\u800c\u4e0d\u662f\u4e0e\u5176\u5bf9\u6297\u3002<\/p>\n<p>\u7528\u4e8e\u7f16\u7801\u4efb\u52a1\u7684 SWE-agent ACI \u5305\u542b\u56db\u4e2a\u4e3b\u8981\u7ec4\u4ef6\uff0c\u6bcf\u4e00\u4e2a\u90fd\u53cd\u6620\u4e86\u5173\u4e8e\u8bed\u8a00\u6a21\u578b\u5728\u83b7\u5f97\u539f\u59cb\u8ba1\u7b97\u673a\u8bbf\u95ee\u6743\u9650\u65f6\u5982\u4f55\u5931\u8d25\u7684\u7279\u5b9a\u6d1e\u5bdf\u3002<\/p>\n<h4>\u641c\u7d22\u4e0e\u5bfc\u822a<\/h4>\n<p>\u641c\u7d22\u7ec4\u4ef6\u7528\u4e13\u95e8\u6784\u5efa\u7684\u5de5\u5177\uff08`find_file`\u3001`search_file`\u3001`search_dir`\uff09\u53d6\u4ee3\u4e86\u6807\u51c6\u7684 `grep` \u548c `find` \u547d\u4ee4\u3002\u5173\u952e\u533a\u522b\u4e0d\u5728\u4e8e\u8bed\u6cd5\uff0c\u800c\u5728\u4e8e\u8f93\u51fa\u7ba1\u7406\u3002\u7ed3\u679c\u88ab\u9650\u5236\u5728 50 \u6761\u4ee5\u5185\u3002\u5982\u679c\u67e5\u8be2\u8d85\u8fc7\u8be5\u9650\u5236\uff0c\u5de5\u5177\u4f1a\u8fd4\u56de\u4e00\u6761\u6d88\u606f\uff0c\u89e3\u91ca\u7ed3\u679c\u8fc7\u591a\u5e76\u63d0\u793a\u667a\u80fd\u4f53\u7ec6\u5316\u641c\u7d22\u3002\u8fd9\u542c\u8d77\u6765\u5fae\u4e0d\u8db3\u9053\uff0c\u4f46\u5728\u5b9e\u8df5\u4e2d\uff0c\u5b83\u662f\u8be5\u8bba\u6587\u4e2d\u6700\u91cd\u8981\u7684\u51b3\u7b56\u4e4b\u4e00\u3002<\/p>\n<p>\u5176\u4e4b\u6240\u4ee5\u91cd\u8981\uff0c\u662f\u56e0\u4e3a\u667a\u80fd\u4f53\u548c\u5904\u4e8e\u8ba4\u77e5\u8d1f\u8377\u4e0b\u7684\u4eba\u7c7b\u4e00\u6837\uff0c\u5f53\u611f\u5230\u4e0d\u786e\u5b9a\u65f6\uff0c\u5f80\u5f80\u4f1a\u4e0d\u65ad\u91cd\u590d\u6b63\u5728\u505a\u7684\u4e8b\u60c5\u3002\u5f53\u4eba\u7c7b\u8ff7\u5931\u5728\u5e9e\u5927\u7684\u4ee3\u7801\u5e93\u4e2d\u65f6\uff0c\u4ed6\u4eec\u4f1a\u641c\u7d22\u5f97\u8d8a\u6765\u8d8a\u5bbd\u6cdb\uff0c\u4ea7\u751f\u8d8a\u6765\u8d8a\u591a\u7684\u566a\u97f3\u3002\u53d7\u9650\u641c\u7d22\u5de5\u5177\u901a\u8fc7\u521b\u9020\u4e00\u4e2a\u201c\u5f3a\u5236\u51fd\u6570\u201d\u4e2d\u65ad\u4e86\u8fd9\u79cd\u6a21\u5f0f\uff1a\u4f60\u4e0d\u80fd\u901a\u8fc7\u6a21\u7cca\u4e0d\u6e05\u6765\u63a8\u8fdb\u5de5\u4f5c\uff0c\u4f60\u5fc5\u987b\u53d8\u5f97\u7cbe\u786e\u3002\u8fd9\u63a8\u52a8\u4e86\u667a\u80fd\u4f53\u91c7\u53d6\u66f4\u6df1\u601d\u719f\u8651\u3001\u66f4\u6709\u9488\u5bf9\u6027\u7684\u884c\u4e3a\u3002<\/p>\n<h4>\u6587\u4ef6\u67e5\u770b\u5668<\/h4>\n<p>\u6587\u4ef6\u67e5\u770b\u5668\u662f\u8bba\u6587\u4e2d\u5173\u4e8e\u8ba4\u77e5\u67b6\u6784\u7684\u6d1e\u5bdf\u53d8\u5f97\u6700\u5177\u4f53\u7684\u5730\u65b9\u3002\u7814\u7a76\u4eba\u5458\u6d4b\u8bd5\u4e86\u591a\u79cd\u67e5\u770b\u5668\u914d\u7f6e\uff0c\u53d1\u73b0\u4e00\u6b21\u663e\u793a 100 \u884c\u662f\u4e00\u4e2a\u201c\u6070\u5230\u597d\u5904\uff08Goldilocks\uff09\u201d\u7684\u6570\u5b57\u3002\u663e\u793a\u884c\u6570\u66f4\u5c11\uff08\u4ed6\u4eec\u6d4b\u8bd5\u4e86 30 \u884c\uff09\u4f1a\u5bfc\u81f4\u667a\u80fd\u4f53\u4e22\u5931\u5468\u56f4\u4ee3\u7801\u7684\u4e0a\u4e0b\u6587\u5e76\u4ea7\u751f\u7f16\u8f91\u9519\u8bef\uff1b\u663e\u793a\u66f4\u591a\u884c\uff08\u6216\u5168\u91cf\u6587\u4ef6\uff09\u5219\u4f1a\u5bfc\u81f4\u667a\u80fd\u4f53\u8ff7\u5931\u6240\u5728\u4f4d\u7f6e\u5e76\u9057\u6f0f\u91cd\u8981\u7ec6\u8282\u3002<\/p>\n<p>\u67e5\u770b\u5668\u662f\u6709\u72b6\u6001\u7684\u3002\u5b83\u5728\u4ea4\u4e92\u8fc7\u7a0b\u4e2d\u4fdd\u6301\u5728\u6587\u4ef6\u4e2d\u7684\u4f4d\u7f6e\u3002\u5173\u952e\u5728\u4e8e\uff0c\u5b83\u5728\u6bcf\u4e2a\u53ef\u89c1\u884c\u524d\u90fd\u6dfb\u52a0\u4e86\u660e\u786e\u7684\u884c\u53f7\u3002\u8fd9\u6700\u540e\u4e00\u4e2a\u7ec6\u8282\u542c\u8d77\u6765\u50cf\u662f\u88c5\u9970\u6027\u7684\uff0c\u5176\u5b9e\u4e0d\u7136\u3002\u5f53\u667a\u80fd\u4f53\u9700\u8981\u53d1\u5e03\u9488\u5bf9\u7b2c 47 \u5230 52 \u884c\u7684\u7f16\u8f91\u547d\u4ee4\u65f6\uff0c\u5b83\u9700\u8981\u80fd\u591f\u76f4\u63a5\u4ece\u89c6\u56fe\u4e2d\u8bfb\u53d6\u8fd9\u4e9b\u6570\u5b57\uff0c\u800c\u4e0d\u662f\u53bb\u6570\u884c\u6570\u6216\u8fdb\u884c\u7b97\u672f\u8fd0\u7b97\u3002\u4ece\u667a\u80fd\u4f53\u7684\u5de5\u4f5c\u8bb0\u5fc6\u4e2d\u79fb\u9664\u8fd9\u9879\u8ba4\u77e5\u4efb\u52a1\uff0c\u4e3a\u5b9e\u9645\u7684\u540e\u7eed\u95ee\u9898\u89e3\u51b3\u817e\u51fa\u4e86\u7a7a\u95f4\u3002<\/p>\n<h4>\u5e26\u6709 Lint \u68c0\u67e5\u7684\u6587\u4ef6\u7f16\u8f91\u5668<\/h4>\n<p>\u6587\u4ef6\u7f16\u8f91\u5668\u7684\u5173\u952e\u521b\u65b0\u662f\u5e26\u6709\u62a4\u680f\u7684\u5373\u65f6\u53cd\u9988\u3002`edit` \u547d\u4ee4\u63a5\u53d7\u8d77\u59cb\u884c\u3001\u7ed3\u675f\u884c\u548c\u66ff\u6362\u6587\u672c\u4f5c\u4e3a\u5355\u4e00\u64cd\u4f5c\u3002\u6bcf\u6b21\u7f16\u8f91\u540e\uff0c\u5de5\u5177\u4f1a\u81ea\u52a8\u5bf9\u4fee\u6539\u540e\u7684\u6587\u4ef6\u8fd0\u884c\u00a0Linter\uff08\u9759\u6001\u4ee3\u7801\u68c0\u67e5\uff09\u5e76\u62a5\u544a\u7ed3\u679c\u3002\u5982\u679c\u7f16\u8f91\u5f15\u5165\u4e86\u8bed\u6cd5\u9519\u8bef\uff0c\u7f16\u8f91\u4f1a\u5728\u5e94\u7528\u524d\u88ab\u62d2\u7edd\uff0c\u667a\u80fd\u4f53\u5c06\u6536\u5230\u4e00\u6761\u6e05\u6670\u7684\u9519\u8bef\u6d88\u606f\uff0c\u540c\u65f6\u663e\u793a\u539f\u59cb\u4ee3\u7801\u548c\u5931\u8d25\u7684\u7f16\u8f91\u5185\u5bb9\u3002<\/p>\n<p>\u8fd9\u95ed\u5408\u4e86\u5728\u539f\u751f\u667a\u80fd\u4f53\u5b9e\u73b0\u4e2d\u5bfc\u81f4\u7ea7\u8054\u6545\u969c\u7684\u53cd\u9988\u5faa\u73af\u3002\u5982\u679c\u6ca1\u6709 Linter\uff0c\u667a\u80fd\u4f53\u53ef\u80fd\u4f1a\u5f15\u5165\u8bed\u6cd5\u9519\u8bef\uff0c\u7136\u540e\u8fd0\u884c\u6d4b\u8bd5\u5957\u4ef6\uff0c\u770b\u5230\u4e00\u4e2a\u4f3c\u4e4e\u4e0d\u76f8\u5173\u7684\u5931\u8d25\uff08\u56e0\u4e3a\u771f\u6b63\u7684\u9519\u8bef\u5728\u522b\u5904\uff09\uff0c\u82b1\u8d39\u591a\u4e2a\u6b65\u9aa4\u8ffd\u67e5\u9519\u8bef\u7684\u95ee\u9898\uff0c\u6700\u540e\u5728\u8ffd\u9010\u5e7b\u5f71\u4e2d\u8017\u5c3d\u5176\u4e0a\u4e0b\u6587\u7a97\u53e3\u3002\u901a\u8fc7\u5c06 Linter \u76f4\u63a5\u96c6\u6210\u5230\u7f16\u8f91\u5668\u4e2d\uff0c\u8bed\u6cd5\u9519\u8bef\u5728\u5f15\u5165\u77ac\u95f4\u5c31\u88ab\u6355\u83b7\uff0c\u4fee\u590d\u5de5\u4f5c\u5728\u95ee\u9898\u4f20\u64ad\u524d\u5c31\u88ab\u5c40\u90e8\u5316\u4e86\u3002<\/p>\n<p>\u5bf9\u6bd4\u667a\u80fd\u4f53\u4f7f\u7528\u539f\u59cb bash \u5de5\u5177\u65f6\u7684\u60c5\u51b5\uff1a\u4f7f\u7528 `sed` \u6216\u8f93\u51fa\u91cd\u5b9a\u5411\u65f6\uff0c\u6ca1\u6709\u96c6\u6210\u53cd\u9988\uff0c\u7f16\u8f91\u9759\u9ed8\u6267\u884c\u3002\u591a\u884c\u66f4\u6539\u9700\u8981\u590d\u6742\u7684\u53c2\u6570\u683c\u5f0f\u5316\uff0c\u6781\u5176\u5bb9\u6613\u51fa\u9519\u3002\u667a\u80fd\u4f53\u53ef\u80fd\u6210\u529f\u8fd0\u884c\u4e86\u547d\u4ee4\u5374\u5f15\u5165\u4e86\u7ec6\u5fae\u7684\u683c\u5f0f\u9519\u8bef\uff08\u672c\u53ef\u4ee5\u88ab Linter \u6355\u83b7\uff09\uff0c\u7136\u540e\u63a5\u4e0b\u6765\u7684\u5341\u4e2a\u6b65\u9aa4\u90fd\u5728\u7591\u60d1\u4e3a\u4ec0\u4e48\u6d4b\u8bd5\u4f1a\u5931\u8d25\u3002<\/p>\n<h4>\u4e0a\u4e0b\u6587\u7ba1\u7406<\/h4>\n<p>\u7b2c\u56db\u4e2a\u7ec4\u4ef6\u89e3\u51b3\u4e86\u4e00\u4e2a\u5728\u957f\u4f1a\u8bdd\u4e2d\u4e0d\u65ad\u6076\u5316\u7684\u95ee\u9898\uff1a\u9648\u65e7\u4e0a\u4e0b\u6587\u7684\u7d2f\u79ef\u3002\u968f\u7740\u667a\u80fd\u4f53\u6267\u884c\u4efb\u52a1\uff0c\u5176\u5386\u53f2\u8bb0\u5f55\u5145\u6ee1\u4e86\u65e7\u7684\u89c2\u5bdf\u7ed3\u679c\u3001\u4e2d\u95f4\u72b6\u6001\u548c\u63a2\u7d22\u6b65\u9aa4\uff0c\u8fd9\u4e9b\u5df2\u4e0d\u518d\u53cd\u6620\u73af\u5883\u7684\u5f53\u524d\u72b6\u6001\u3002\u6240\u6709\u8fd9\u4e9b\u5386\u53f2\u8bb0\u5f55\u90fd\u5360\u636e\u4e86\u4e0a\u4e0b\u6587\u7a97\u53e3\u7684\u7a7a\u95f4\uff0c\u5e76\u53ef\u80fd\u901a\u8fc7\u63d0\u4f9b\u8fc7\u65f6\u4fe1\u606f\u6765\u8bef\u5bfc\u667a\u80fd\u4f53\u3002<\/p>\n<p>ACI \u7684\u4e0a\u4e0b\u6587\u7ba1\u7406\u7cfb\u7edf\u5c06\u8f83\u65e7\u7684\u89c2\u5bdf\u7ed3\u679c\uff08\u6700\u540e\u4e94\u8f6e\u4e4b\u524d\u7684\uff09\u6298\u53e0\u6210\u5355\u884c\u6458\u8981\u3002\u8fd9\u4f7f\u6d3b\u8dc3\u4e0a\u4e0b\u6587\u805a\u7126\u4e8e\u6700\u8fd1\u7684\u76f8\u5173\u4fe1\u606f\uff0c\u540c\u65f6\u4fdd\u7559\u4e86\u6574\u4f53\u8f68\u8ff9\u7684\u538b\u7f29\u8bb0\u5f55\u3002\u667a\u80fd\u4f53\u59cb\u7ec8\u80fd\u770b\u5230\u5b83\u6700\u8fd1\u505a\u4e86\u4ec0\u4e48\u4ee5\u53ca\u5f53\u524d\u72b6\u6001\u662f\u4ec0\u4e48\uff0c\u800c\u4e0d\u4f1a\u88ab\u5b83\u8fd0\u884c\u8fc7\u7684\u6bcf\u4e2a\u547d\u4ee4\u7684\u5b8c\u6574\u672a\u538b\u7f29\u5386\u53f2\u6240\u6df9\u6ca1\u3002<\/p>\n<h4>\u57fa\u51c6\u6d4b\u8bd5\u7ed3\u679c\u53ca\u5176\u771f\u5b9e\u542b\u4e49<\/h4>\n<p>SWE-agent \u8bba\u6587\u9488\u5bf9 SWE-bench \u6570\u636e\u96c6\uff08\u6765\u81ea\u6d41\u884c Python \u4ed3\u5e93\u7684\u771f\u5b9e GitHub Issue \u96c6\u5408\uff09\u5bf9\u5176 ACI \u8fdb\u884c\u4e86\u57fa\u51c6\u6d4b\u8bd5\u3002\u4efb\u52a1\u662f\u63a5\u6536\u4e00\u4e2a\u81ea\u7136\u8bed\u8a00\u9519\u8bef\u62a5\u544a\u5e76\u4ea7\u51fa\u4e00\u4e2a\u89e3\u51b3\u5b83\u7684\u4ee3\u7801\u8865\u4e01\u3002\u8fd9\u662f\u4e00\u4e2a\u8270\u5de8\u7684\u5b9e\u8df5\u4efb\u52a1\uff0c\u9700\u8981\u5bfc\u822a\u4e0d\u719f\u6089\u7684\u4ee3\u7801\u5e93\u3001\u7406\u89e3\u9519\u8bef\u6d88\u606f\u3001\u7f16\u5199\u6b63\u786e\u7684\u4ee3\u7801\u5e76\u9a8c\u8bc1\u4fee\u590d\u3002<\/p>\n<p>\u4f7f\u7528\u5e26\u6709\u6807\u51c6 bash shell \u63a5\u53e3\u7684 GPT-4\uff0c\u7cfb\u7edf\u89e3\u51b3\u4e86 3.97% \u7684\u95ee\u9898\u3002\u4f7f\u7528\u5e26\u6709\u4e13\u95e8\u6784\u5efa\u7684 ACI \u7684 GPT-4\uff0c\u7cfb\u7edf\u89e3\u51b3\u4e86 12.47% \u7684\u95ee\u9898\u3002\u8fd9\u5c31\u662f\u4e4b\u524d\u63d0\u5230\u7684 64% \u7684\u76f8\u5bf9\u63d0\u5347\uff0c\u4e14\u8fd9\u5b8c\u5168\u6765\u81ea\u4e8e\u63a5\u53e3\u8bbe\u8ba1\u3002<\/p>\n<p>\u7814\u7a76\u4eba\u5458\u8fd8\u8fdb\u884c\u4e86\u6d88\u878d\u5b9e\u9a8c\uff0c\u6bcf\u6b21\u79fb\u9664\u4e00\u4e2a\u7ec4\u4ef6\u4ee5\u9694\u79bb\u6bcf\u4e2a\u8bbe\u8ba1\u51b3\u7b56\u7684\u8d21\u732e\u3002Linter \u96c6\u6210\u59cb\u7ec8\u662f\u6760\u6746\u4f5c\u7528\u6700\u9ad8\u7684\u7ec4\u4ef6\u4e4b\u4e00\u3002\u53d7\u9650\u641c\u7d22\u5bf9\u4e8e\u9632\u6b62\u4e0a\u4e0b\u6587\u6cdb\u6ee5\u81f3\u5173\u91cd\u8981\u3002\u5e26\u6709\u884c\u53f7\u7684\u6709\u72b6\u6001\u6587\u4ef6\u67e5\u770b\u5668\u8868\u73b0\u663e\u8457\u4f18\u4e8e\u539f\u59cb `cat` \u547d\u4ee4\u548c\u66f4\u7b80\u5355\u7684\u67e5\u770b\u5668\u8bbe\u8ba1\u3002<\/p>\n<p>\u6027\u80fd\u5dee\u5f02\u4e0d\u5728\u4e8e\u6a21\u578b\u7684\u667a\u80fd\u7a0b\u5ea6\uff0c\u800c\u5728\u4e8e\u8ba4\u77e5\u8d1f\u8377\u7ba1\u7406\u3002ACI \u51cf\u5c11\u4e86\u6a21\u578b\u8ddf\u8e2a\u72b6\u6001\u6240\u9700\u7684\u5de5\u4f5c\uff0c\u4e3a\u771f\u6b63\u91cd\u8981\u7684\u5de5\u4f5c\u7559\u51fa\u4e86\u7a7a\u95f4\u3002<\/p>\n<p>\u5176\u5f71\u54cd\u8fdc\u8d85\u7f16\u7801\u667a\u80fd\u4f53\u3002\u4efb\u4f55\u957f\u7a0b\uff08long-horizon\uff09\u667a\u80fd\u4f53\u4efb\u52a1\u90fd\u6d89\u53ca\u76f8\u540c\u7684\u6839\u672c\u6311\u6218\uff1a\u5728\u5e9e\u5927\u7684\u4fe1\u606f\u7a7a\u95f4\u4e2d\u5bfc\u822a\u3001\u5728\u591a\u4e2a\u6b65\u9aa4\u4e2d\u4fdd\u6301\u8fde\u8d2f\u7684\u72b6\u6001\u3001\u6355\u83b7\u9519\u8bef\u5e76\u4ece\u4e2d\u6062\u590d\uff0c\u4ee5\u53ca\u7ba1\u7406\u6709\u9650\u7684\u4e0a\u4e0b\u6587\u7a97\u53e3\u6ce8\u610f\u529b\u8d44\u6e90\u3002ACI \u7684\u8bbe\u8ba1\u539f\u5219\u662f\u666e\u9002\u7684\uff0c\u5177\u4f53\u7684\u5de5\u5177\u4f1a\u53d8\uff0c\u95ee\u9898\u7684\u5e95\u5c42\u67b6\u6784\u4e0d\u4f1a\u53d8\u3002<\/p>\n<h3>Anthropic \u7684Harness\u5de5\u7a0b\uff08\u957f\u7a0b\u8fd0\u884c\u667a\u80fd\u4f53\u95ee\u9898\uff09<\/h3>\n<h4>\u4e3a\u4ec0\u4e48\u4e0a\u4e0b\u6587\u7a97\u53e3\u8fb9\u754c\u662f\u96be\u9898<\/h4>\n<p>SWE-agent \u8bba\u6587\u89e3\u51b3\u4e86\u5982\u4f55\u4e3a\u5355\u6b21\u667a\u80fd\u4f53\u4f1a\u8bdd\u8bbe\u8ba1\u63a5\u53e3\u3002\u800c Anthropic \u7684\u5de5\u7a0b\u56e2\u961f\u5728\u5f00\u53d1 Claude Agent SDK \u548c Claude Code \u65f6\u9047\u5230\u4e86\u53e6\u4e00\u4e2a\u95ee\u9898\uff1a\u5f53\u4efb\u52a1\u5927\u5230\u65e0\u6cd5\u5728\u5355\u4e2a\u4e0a\u4e0b\u6587\u7a97\u53e3\u4e2d\u5b8c\u6210\u65f6\uff0c\u4f1a\u53d1\u751f\u4ec0\u4e48\uff1f<\/p>\n<p>\u8fd9\u5e76\u4e0d\u662f\u4e00\u4e2a\u51b7\u95e8\u7684\u8fb9\u7f18\u6848\u4f8b\u3002\u5927\u591a\u6570\u771f\u5b9e\u7684\u8f6f\u4ef6\u9879\u76ee\u90fd\u592a\u5927\uff0c\u65e0\u6cd5\u653e\u5165\u4efb\u4f55\u4e0a\u4e0b\u6587\u7a97\u53e3\u3002\u4e00\u4e2a\u751f\u4ea7\u73af\u5883\u7684 Web \u5e94\u7528\u62e5\u6709\u6570\u767e\u4e2a\u6587\u4ef6\u3001\u6570\u5343\u4e2a\u51fd\u6570\u3001\u6d4b\u8bd5\u5957\u4ef6\u3001\u914d\u7f6e\u3001\u6587\u6863\u548c\u4f9d\u8d56\u9879\u3002\u5373\u4f7f\u62e5\u6709 200K Token \u7684\u4e0a\u4e0b\u6587\u7a97\u53e3\uff0c\u4f60\u4e5f\u65e0\u6cd5\u540c\u65f6\u5728\u5927\u8111\u4e2d\u4fdd\u7559\u6574\u4e2a\u9879\u76ee\u3002\u4eba\u7c7b\u5de5\u7a0b\u5e08\u901a\u8fc7\u5916\u90e8\u8bb0\u5fc6\u3001\u6587\u6863\u3001\u7248\u672c\u63a7\u5236\u4ee5\u53ca\u5728\u4ee3\u7801\u5e93\u4e2d\u5de5\u4f5c\u6570\u5468\u751a\u81f3\u6570\u6708\u79ef\u7d2f\u7684\u7406\u89e3\u6765\u89e3\u51b3\u8fd9\u4e2a\u95ee\u9898\u3002\u800c\u4e00\u4e2a\u5f00\u542f\u65b0\u4f1a\u8bdd\u7684\u667a\u80fd\u4f53\u5b8c\u5168\u6ca1\u6709\u8fd9\u4e9b\u3002<\/p>\n<p>\u539f\u751f\u7684\u89e3\u51b3\u65b9\u6848\u662f\u538b\u7f29\uff0c\u5b83\u5728\u67d0\u79cd\u7a0b\u5ea6\u4e0a\u6709\u6548\u3002Claude Agent SDK \u5305\u542b\u538b\u7f29\u529f\u80fd\uff0c\u5f53\u7a97\u53e3\u586b\u6ee1\u65f6\u4f1a\u603b\u7ed3\u65e7\u7684\u4e0a\u4e0b\u6587\u3002\u4f46\u4ec5\u9760\u538b\u7f29\u662f\u4e0d\u591f\u7684\u3002Anthropic \u7684\u5185\u90e8\u5b9e\u9a8c\u8868\u660e\uff0c\u5373\u4f7f\u6709\u538b\u7f29\uff0c\u50cf Opus 4.5 \u8fd9\u6837\u5904\u4e8e\u524d\u6cbf\u7684\u7f16\u7801\u6a21\u578b\uff0c\u5982\u679c\u5728\u591a\u4e2a\u4e0a\u4e0b\u6587\u7a97\u53e3\u4e2d\u5faa\u73af\u8fd0\u884c\uff0c\u4ecd\u7136\u65e0\u6cd5\u6839\u636e\u4e00\u4e2a\u9ad8\u5c42\u6b21\u63d0\u793a\u8bcd\u6784\u5efa\u51fa\u4e00\u4e2a\u751f\u4ea7\u7ea7\u8d28\u91cf\u7684 Web \u5e94\u7528\u3002<\/p>\n<p>\u5931\u8d25\u4e3b\u8981\u96c6\u4e2d\u5728\u4e24\u79cd\u6a21\u5f0f\u4e0a\uff0c\u8fd9\u4e24\u79cd\u6a21\u5f0f\u90fd\u6781\u5177\u542f\u53d1\u6027\u3002<\/p>\n<p>\u7b2c\u4e00\u79cd\u5931\u8d25\u6a21\u5f0f\u662f<strong>\u8bd5\u56fe\u4e00\u6b21\u505a\u592a\u591a\u4e8b\u60c5<\/strong>\u3002\u5f53\u6536\u5230\u201c\u6784\u5efa\u4e00\u4e2a claude.ai \u7684\u514b\u9686\u7248\u201d\u4e4b\u7c7b\u7684\u63d0\u793a\u8bcd\u65f6\uff0c\u667a\u80fd\u4f53\u4f1a\u5c1d\u8bd5\u201c\u4e00\u7ad9\u5f0f\u201d\u5b8c\u6210\u6574\u4e2a\u5e94\u7528\u3002\u5b83\u4f1a\u5728\u4e0d\u5b8c\u6210\u6216\u4e0d\u6d4b\u8bd5\u4efb\u4f55\u529f\u80fd\u7684\u60c5\u51b5\u4e0b\u5f00\u59cb\u5b9e\u73b0\u4e00\u4e2a\u53c8\u4e00\u4e2a\u529f\u80fd\uff0c\u5728\u5b9e\u73b0\u8fc7\u7a0b\u4e2d\u8017\u5c3d\u4e0a\u4e0b\u6587\u7a97\u53e3\uff0c\u5bfc\u81f4\u4e0b\u4e00\u6b21\u4f1a\u8bdd\u5f00\u59cb\u65f6\u9762\u5bf9\u7684\u662f\u4e00\u4e2a\u53ea\u5b9e\u73b0\u4e86\u4e00\u534a\u7684\u5e94\u7528\u3001\u6ca1\u6709\u4efb\u4f55\u5df2\u5b8c\u6210\u5de5\u4f5c\u7684\u6587\u6863\uff0c\u4e5f\u6ca1\u6709\u660e\u786e\u7684\u4ee3\u7801\u72b6\u6001\u6307\u793a\u3002\u4e0b\u4e00\u4e2a\u667a\u80fd\u4f53\u5b9e\u4f8b\u5c06\u82b1\u8d39\u5927\u90e8\u5206\u4e0a\u4e0b\u6587\u9884\u7b97\u53bb\u7406\u89e3\u8fd9\u4e00\u56e2\u7cdf\uff0c\u800c\u4e0d\u662f\u53d6\u5f97\u8fdb\u5c55\u3002<\/p>\n<p>\u7b2c\u4e8c\u79cd\u5931\u8d25\u6a21\u5f0f\u51fa\u73b0\u5728\u9879\u76ee\u540e\u671f\u3002\u5728\u6784\u5efa\u4e86\u4e00\u4e9b\u529f\u80fd\u540e\uff0c\u968f\u540e\u7684\u667a\u80fd\u4f53\u5b9e\u4f8b\u4f1a\u89c2\u5bdf\u56db\u5468\uff0c\u770b\u5230\u5df2\u7ecf\u53d6\u5f97\u4e86\u8fdb\u5c55\uff0c\u4fbf<strong>\u65ad\u5b9a\u5de5\u4f5c\u5df2\u7ecf\u5b8c\u6210<\/strong>\u3002\u5b83\u4f1a\u9488\u5bf9\u4e00\u4e2a\u4ec5\u5b8c\u6210\u90e8\u5206\u7684\u5e94\u7528\u5ba3\u5e03\u80dc\u5229\u5e76\u505c\u6b62\u5de5\u4f5c\u3002\u8fd9\u4e0d\u662f\u611a\u8822\uff0c\u800c\u662f\u57fa\u4e8e\u4e0d\u5b8c\u6574\u4fe1\u606f\u7684\u5408\u7406\u63a8\u8bba\u3002\u667a\u80fd\u4f53\u6ca1\u6709\u7ed3\u6784\u5316\u7684\u65b9\u5f0f\u6765\u4e86\u89e3\u8be5\u9879\u76ee\u771f\u6b63\u7684\u201c\u5b8c\u6210\u201d\u610f\u5473\u7740\u4ec0\u4e48\u3002<\/p>\n<p>\u8fd9\u4e24\u4e2a\u5931\u8d25\u90fd\u6709\u4e00\u4e2a\u5171\u540c\u7684\u6839\u6e90\uff1a\u667a\u80fd\u4f53\u5bf9\u9879\u76ee\u72b6\u6001\u6ca1\u6709\u6301\u4e45\u7684\u3001\u7ed3\u6784\u5316\u7684\u7406\u89e3\uff0c\u8fd9\u79cd\u7406\u89e3\u65e0\u6cd5\u8de8\u8d8a\u4e0a\u4e0b\u6587\u7a97\u53e3\u8fb9\u754c\u5e76\u5f15\u5bfc\u672a\u6765\u7684\u4f1a\u8bdd\u3002<\/p>\n<h4>\u53cc\u667a\u80fd\u4f53\u67b6\u6784\uff1a\u521d\u59cb\u5316\u667a\u80fd\u4f53\u4e0e\u7f16\u7801\u667a\u80fd\u4f53<\/h4>\n<p>Anthropic \u7684\u89e3\u51b3\u65b9\u6848\u662f\u4e00\u4e2a\u7531\u4e24\u90e8\u5206\u7ec4\u6210\u7684\u67b6\u6784\uff0c\u8fd9\u5df2\u6210\u4e3a\u4e25\u8083\u56e2\u961f\u5904\u7406\u957f\u7a0b\u667a\u80fd\u4f53\u5316\u5de5\u4f5c\u7684\u6a21\u677f\u3002<\/p>\n<p>\u7b2c\u4e00\u90e8\u5206\u662f<strong>\u521d\u59cb\u5316\u667a\u80fd\u4f53<\/strong>\u3002\u8fd9\u662f\u4e00\u4e2a\u4e13\u95e8\u7684\u9996\u6b21\u4f1a\u8bdd\uff0c\u5e26\u6709\u72ec\u7279\u7684\u7cfb\u7edf\u63d0\u793a\u8bcd\uff0c\u5176\u5168\u90e8\u76ee\u7684\u662f\u8bbe\u7f6e\u672a\u6765\u6240\u6709\u7f16\u7801\u667a\u80fd\u4f53\u90fd\u5c06\u5728\u5176\u4e2d\u64cd\u4f5c\u7684\u73af\u5883\u3002\u5b83\u4e0d\u7f16\u5199\u529f\u80fd\u4ee3\u7801\uff0c\u800c\u662f\u521b\u5efa\u8ba9\u8de8\u591a\u4e2a\u540e\u7eed\u4f1a\u8bdd\u7684\u529f\u80fd\u5f00\u53d1\u6210\u4e3a\u53ef\u80fd\u7684\u811a\u624b\u67b6\u3002<\/p>\n<p>\u521d\u59cb\u5316\u667a\u80fd\u4f53\u4ea7\u51fa\u4e09\u4e2a\u5173\u952e\u8f93\u51fa\u3002\u9996\u5148\uff0c\u5b83\u521b\u5efa\u4e00\u4e2a\u00a0`init.sh` \u811a\u672c\uff0c\u53ef\u4ee5\u53ef\u9760\u5730\u542f\u52a8\u5f00\u53d1\u73af\u5883\u3002\u8fd9\u542c\u8d77\u6765\u5f88\u5e73\u51e1\uff0c\u4f46\u5b83\u5177\u6709\u5de8\u5927\u7684\u6760\u6746\u4f5c\u7528\u3002\u6b64\u540e\u7684\u6bcf\u4e00\u6b21\u7f16\u7801\u667a\u80fd\u4f53\u4f1a\u8bdd\u90fd\u53ef\u4ee5\u901a\u8fc7\u8fd0\u884c `init.sh` \u5f00\u59cb\uff0c\u800c\u4e0d\u662f\u82b1\u8d39 Token \u53bb\u6478\u7d22\u5982\u4f55\u542f\u52a8\u670d\u52a1\u5668\u3001\u8bbe\u7f6e\u6570\u636e\u5e93\u4ee5\u53ca\u8ba9\u5e94\u7528\u8fdb\u5165\u53ef\u6d4b\u8bd5\u72b6\u6001\u3002\u5728\u6bcf\u6b21\u4f1a\u8bdd\u4e2d\u8282\u7701\u7684\u8fd9\u4e9b\u5f00\u9500\u7d2f\u79ef\u8d77\u6765\u975e\u5e38\u53ef\u89c2\u3002<\/p>\n<p>\u5176\u6b21\uff0c\u521d\u59cb\u5316\u667a\u80fd\u4f53\u521b\u5efa\u4e00\u4e2a<strong>\u5168\u9762\u7684\u529f\u80fd\u5217\u8868\u6587\u4ef6<\/strong>\u3002\u5728 Anthropic \u5185\u90e8\u8fdb\u884c\u7684 claude.ai \u514b\u9686\u5b9e\u9a8c\u4e2d\uff0c\u8fd9\u610f\u5473\u7740\u8d85\u8fc7 200 \u4e2a\u5177\u4f53\u7684\u3001\u7aef\u5230\u7aef\u7684\u529f\u80fd\u63cf\u8ff0\uff0c\u4f8b\u5982\u201c\u7528\u6237\u53ef\u4ee5\u6253\u5f00\u65b0\u804a\u5929\u3001\u8f93\u5165\u67e5\u8be2\u3001\u6309\u56de\u8f66\u5e76\u770b\u5230 AI \u54cd\u5e94\u201d\u3002\u6bcf\u4e2a\u529f\u80fd\u6700\u521d\u90fd\u6807\u8bb0\u4e3a\u5931\u8d25\u3002\u8be5\u6587\u4ef6\u4f5c\u4e3a\u9879\u76ee\u7684\u201c\u552f\u4e00\u771f\u76f8\u201d\u3002\u5f00\u542f\u65b0\u4f1a\u8bdd\u7684\u7f16\u7801\u667a\u80fd\u4f53\u8bfb\u53d6\u6b64\u6587\u4ef6\uff0c\u7acb\u5373\u5c31\u80fd\u786e\u5b9a\u54ea\u4e9b\u5df2\u7ecf\u6784\u5efa\u3001\u54ea\u4e9b\u5c1a\u672a\u6784\u5efa\u3002\u5b83\u4e0d\u4f1a\u56e0\u4e3a\u770b\u5230\u4e00\u4e9b\u4ee3\u7801\u5c31\u65ad\u5b9a\u5de5\u4f5c\u5df2\u5b8c\u6210\uff0c\u529f\u80fd\u5217\u8868\u4f1a\u544a\u8bc9\u5b83\u771f\u76f8\u3002<\/p>\n<p>\u7b2c\u4e09\uff0c\u521d\u59cb\u5316\u667a\u80fd\u4f53\u521b\u5efa\u4e00\u4e2a `claude-progress.txt` \u6587\u4ef6\u5e76\u8fdb\u884c\u521d\u59cb\u00a0git \u63d0\u4ea4\u3002\u8fdb\u5ea6\u6587\u4ef6\u662f\u4e00\u4e2a\u4eba\u7c7b\u53ef\u8bfb\u7684\u65e5\u5fd7\uff0c\u667a\u80fd\u4f53\u5728\u6bcf\u6b21\u4f1a\u8bdd\u7ed3\u675f\u65f6\u66f4\u65b0\u5b83\uff0c\u8bb0\u5f55\u5b83\u4eec\u5904\u7406\u4e86\u4ec0\u4e48\u3001\u5b8c\u6210\u4e86\u4ec0\u4e48\u4ee5\u53ca\u7559\u4e0b\u7684\u72b6\u6001\u3002\u7ed3\u5408 git \u5386\u53f2\u8bb0\u5f55\uff0c\u8fd9\u8ba9\u672a\u6765\u7684\u6bcf\u4e2a\u7f16\u7801\u667a\u80fd\u4f53\u90fd\u80fd\u5feb\u901f\u5b9a\u4f4d\uff0c\u800c\u65e0\u9700\u5728\u201c\u8003\u53e4\u201d\u4e0a\u8017\u8d39\u4e0a\u4e0b\u6587\u9884\u7b97\u3002<\/p>\n<p>\u7b2c\u4e8c\u90e8\u5206\u662f<strong>\u7f16\u7801\u667a\u80fd\u4f53<\/strong>\u3002\u521d\u59cb\u5316\u4e4b\u540e\u7684\u6bcf\u4e00\u6b21\u4f1a\u8bdd\u90fd\u4f7f\u7528\u4e0d\u540c\u7684\u63d0\u793a\u8bcd\uff1a\u4e00\u6b21\u53ea\u5904\u7406\u4e00\u4e2a\u529f\u80fd\uff0c\u4fdd\u6301\u73af\u5883\u72b6\u6001\u5e72\u51c0\uff0c\u5e76\u5728\u4f1a\u8bdd\u7ed3\u675f\u524d\u66f4\u65b0\u8fdb\u5ea6\u6587\u4ef6\u548c git \u5386\u53f2\u3002\u589e\u91cf\u7684\u8fdb\u5c55\uff0c\u6587\u6863\u5316\u7684\u72b6\u6001\uff0c\u5e72\u51c0\u7684\u4ea4\u63a5\u3002<\/p>\n<h4>\u529f\u80fd\u5217\u8868\u4f5c\u4e3a\u8ba4\u77e5\u951a\u70b9<\/h4>\n<p>\u529f\u80fd\u5217\u8868\u503c\u5f97\u7279\u522b\u5173\u6ce8\uff0c\u56e0\u4e3a\u5b83\u89e3\u51b3\u4e86\u4e00\u4e2a\u6781\u6613\u88ab\u4f4e\u4f30\u7684\u95ee\u9898\u3002\u5982\u679c\u6ca1\u6709\u5b83\uff0c\u5728\u590d\u6742\u4ee3\u7801\u5e93\u4e2d\u64cd\u4f5c\u7684\u667a\u80fd\u4f53\u5fc5\u987b\u4ece\u4ee3\u7801\u672c\u8eab\u63a8\u65ad\u9879\u76ee\u7684\u5b8c\u6210\u60c5\u51b5\u3002\u8fd9\u79cd\u63a8\u8bba\u662f\u4e0d\u53ef\u9760\u7684\u3002\u53ef\u80fd\u5b58\u5728\u4e0d\u53ef\u7528\u7684\u4ee3\u7801\uff0c\u4e5f\u53ef\u80fd\u5b58\u5728\u4e0d\u5b8c\u6574\u7684\u529f\u80fd\u3002\u667a\u80fd\u4f53\u9605\u8bfb\u4ee3\u7801\u5e76\u63a8\u7406\u5df2\u5b8c\u6210\u7684\u5de5\u4f5c\uff0c\u5176\u51fa\u9519\u9891\u7387\u8db3\u4ee5\u6784\u6210\u4e25\u91cd\u95ee\u9898\u3002<\/p>\n<p>\u529f\u80fd\u5217\u8868\u4f7f\u5b8c\u6210\u60c5\u51b5\u53d8\u5f97\u660e\u786e\u4e14\u65e0\u6b67\u4e49\u3002\u6bcf\u4e2a\u529f\u80fd\u90fd\u6709\u4e00\u4e2a `passes` \u5b57\u6bb5\uff0c\u8981\u4e48\u4e3a `true`\uff0c\u8981\u4e48\u4e3a `false`\u3002\u667a\u80fd\u4f53\u8981\u4e48\u5728\u9a8c\u8bc1\u529f\u80fd\u7aef\u5230\u7aef\u8fd0\u884c\u540e\u66f4\u65b0\u6b64\u5b57\u6bb5\uff0c\u8981\u4e48\u4e0d\u66f4\u65b0\u3002\u8fd9\u91cc\u6ca1\u6709\u6b67\u4e49\uff0c\u4e0d\u9700\u8981\u63a8\u8bba\u3002\u771f\u76f8\u5c31\u5728\u6587\u4ef6\u91cc\u3002<\/p>\n<p>Anthropic \u523b\u610f\u51b3\u5b9a\u5c06\u6b64\u5217\u8868\u5b58\u50a8\u4e3a JSON \u800c\u4e0d\u662f Markdown\u3002\u539f\u56e0\u5728\u4e8e\u884c\u4e3a\u5b66\u3002\u7ecf\u9a8c\u8868\u660e\uff0c\u4e0e Markdown \u6587\u4ef6\u76f8\u6bd4\uff0c\u6a21\u578b\u4e0d\u592a\u53ef\u80fd\u4e0d\u6070\u5f53\u5730\u4fee\u6539\u6216\u8986\u76d6 JSON \u6587\u4ef6\u3002JSON \u5177\u6709\u4e25\u5bc6\u7684\u7ed3\u6784\uff0c\u80fd\u591f\u62b5\u5fa1\u968f\u610f\u7684\u7f16\u8f91\u3002\u8fd9\u662f\u4e00\u4e2a\u4ea7\u751f\u771f\u5b9e\u540e\u679c\u7684\u5c0f\u7ec6\u8282\uff1a\u4f60\u5e0c\u671b\u529f\u80fd\u5217\u8868\u662f\u667a\u80fd\u4f53\u8c28\u614e\u66f4\u65b0\u7684\u4e1c\u897f\uff0c\u800c\u4e0d\u662f\u5b83\u4eec\u5174\u8d77\u65f6\u5c31\u968f\u624b\u6539\u5199\u7684\u4e1c\u897f\u3002<\/p>\n<div>\n<div class=\"hljs-cto\">\n<div class=\"hljs-cto\"><button class=\"copy_btn disable\" data-clipboard-target=\"#code_id_0\">\u590d\u5236<\/button><\/p>\n<div class=\"code-toolbar\">\n<pre class=\"has-pre-numbering language-plain\" tabindex=\"0\"><code class=\"language-plain\">{ \"category\": \"functional\", \"description\": \"New chat button creates a fresh conversation\", \"steps\": [ \"Navigate to main interface\", \"Click the 'New Chat' button\", \"Verify a new conversation is created\", \"Check that chat area shows welcome state\", \"Verify conversation appears in sidebar\" ], \"passes\": false }<\/code><\/pre>\n<ul id=\"code_id_0\" class=\"pre-numbering\">\n<li>1.<\/li>\n<\/ul>\n<div class=\"toolbar\"><\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<p>\u4f34\u968f\u6b64\u683c\u5f0f\u7684\u6307\u4ee4\u662f\u660e\u786e\u7684\uff1a\u79fb\u9664\u6216\u7f16\u8f91\u6d4b\u8bd5\u662f\u4e0d\u53ef\u63a5\u53d7\u7684\uff0c\u56e0\u4e3a\u8fd9\u53ef\u80fd\u5bfc\u81f4\u529f\u80fd\u7f3a\u5931\u6216\u4ea7\u751f Bug\u3002\u4f60\u63d0\u793a\u6a21\u578b\u5c06\u6b64\u6587\u4ef6\u89c6\u4e3a\u795e\u5723\u4e0d\u53ef\u4fb5\u72af\u7684\u3002JSON \u7ed3\u6784\u4ece\u67b6\u6784\u4e0a\u5f3a\u5316\u4e86\u8fd9\u4e00\u6307\u4ee4\u3002<\/p>\n<h4>\u589e\u91cf\u8fdb\u5c55\u4e0e\u5e72\u51c0\u72b6\u6001\u8981\u6c42<\/h4>\n<p>\u591a\u4f1a\u8bdd\u667a\u80fd\u4f53\u5de5\u4f5c\u4e2d\uff0c\u6700\u96be\u7684\u95ee\u9898\u4e4b\u4e00\u662f\u786e\u4fdd\u6bcf\u6b21\u4f1a\u8bdd\u7ed3\u675f\u65f6\u7684\u72b6\u6001\u80fd\u8ba9\u4e0b\u6b21\u4f1a\u8bdd\u5b89\u5168\u5730\u5728\u5176\u57fa\u7840\u4e0a\u6784\u5efa\u3002\u5982\u679c\u6ca1\u6709\u660e\u786e\u7684\u5f3a\u5236\u6267\u884c\uff0c\u667a\u80fd\u4f53\u5f80\u5f80\u4f1a\u5728\u4e0a\u4e0b\u6587\u7a97\u53e3\u586b\u6ee1\u65f6\uff0c\u5c06\u5de5\u4f5c\u7559\u5728\u4efb\u4f55\u53ef\u80fd\u7684\u72b6\u6001\uff1a\u5b9e\u73b0\u4e86\u4e00\u534a\u7684\u529f\u80fd\u3001\u635f\u574f\u7684\u6d4b\u8bd5\u3001\u672a\u8bb0\u5f55\u7684\u66f4\u6539\u3002\u4e0b\u4e00\u4e2a\u667a\u80fd\u4f53\u7ee7\u627f\u4e86\u8fd9\u4e00\u56e2\u7cdf\u3002<\/p>\n<p>Anthropic \u7684\u89e3\u51b3\u65b9\u6848\u662f\u5c06\u201c\u5e72\u51c0\u72b6\u6001\u201d\u4f5c\u4e3a\u4e00\u7b49\u516c\u6c11\u7684\u8981\u6c42\uff0c\u800c\u975e\u201c\u6700\u597d\u80fd\u6709\u201d\u3002\u6bcf\u6b21\u7f16\u7801\u667a\u80fd\u4f53\u4f1a\u8bdd\u90fd\u4ee5 git \u63d0\u4ea4\uff08\u5e26\u6709\u63cf\u8ff0\u6027\u6d88\u606f\uff09\u3001\u8fdb\u5ea6\u6587\u4ef6\u66f4\u65b0\u4ee5\u53ca\uff08\u5982\u679c\u9700\u8981\u7684\u8bdd\uff09\u6062\u590d\u5230\u5de5\u4f5c\u72b6\u6001\u7ed3\u675f\u3002\u6240\u8c13\u7684\u201c\u5e72\u51c0\u72b6\u6001\u201d\uff0c\u662f\u6307\u4ee3\u7801\u5e94\u5f53\u9002\u5408\u5408\u5e76\u5230\u4e3b\u5206\u652f\uff1a\u6ca1\u6709\u91cd\u5927 Bug\u3001\u6587\u6863\u5b8c\u5907\u3001\u5904\u4e8e\u4e00\u4e2a\u5f00\u53d1\u8005\u53ef\u4ee5\u5408\u7406\u5730\u5f00\u59cb\u65b0\u529f\u80fd\u800c\u65e0\u9700\u5148\u6e05\u7406\u524d\u4efb\u7559\u4e0b\u7684\u70c2\u644a\u5b50\u7684\u72b6\u6001\u3002<\/p>\n<p>Git \u63d0\u4ea4\u4e0d\u4ec5\u662f\u4e00\u4e2a\u68c0\u67e5\u70b9\uff0c\u5b83\u8fd8\u662f\u4e00\u79cd\u6062\u590d\u673a\u5236\u3002\u5f53\u667a\u80fd\u4f53\u505a\u51fa\u7684\u66f4\u6539\u7834\u574f\u4e86\u67d0\u4e9b\u4e1c\u897f\u65f6\uff0c\u5b83\u53ef\u4ee5\u5229\u7528 git \u6062\u590d\u5230\u4e0a\u4e00\u4e2a\u5df2\u77e5\u7684\u826f\u597d\u72b6\u6001\u5e76\u91cd\u8bd5\u3002\u8fd9\u5c31\u662f\u4eba\u7c7b\u5de5\u7a0b\u5e08\u7684\u5de5\u4f5c\u65b9\u5f0f\uff0c\u4e8b\u5b9e\u8bc1\u660e\uff0c\u8fd9\u5bf9\u667a\u80fd\u4f53\u4e5f\u662f\u5b8c\u5168\u6b63\u786e\u7684\u7eaa\u5f8b\u3002\u7248\u672c\u63a7\u5236\u662f\u8ba4\u77e5\u811a\u624b\u67b6\uff0c\u800c\u4e0d\u4ec5\u4ec5\u662f\u6e90\u7801\u7ba1\u7406\u3002<\/p>\n<h4>\u6d4b\u8bd5\uff1a\u6ca1\u4eba\u559c\u6b22\u8c08\u8bba\u7684\u5931\u8d25\u6a21\u5f0f<\/h4>\n<p>Anthropic \u8bb0\u5f55\u4e86\u4e00\u79cd\u5728\u51e0\u4e4e\u6240\u6709\u4e25\u8083\u7684\u667a\u80fd\u4f53\u5316\u7f16\u7801\u9879\u76ee\u4e2d\u90fd\u4f1a\u51fa\u73b0\u7684\u5931\u8d25\u6a21\u5f0f\uff1a<strong>\u667a\u80fd\u4f53\u5728\u6ca1\u6709\u8fdb\u884c\u6b63\u786e\u7684\u7aef\u5230\u7aef\u9a8c\u8bc1\u7684\u60c5\u51b5\u4e0b\u5c31\u5c06\u529f\u80fd\u6807\u8bb0\u4e3a\u5df2\u5b8c\u6210<\/strong>\u3002\u667a\u80fd\u4f53\u4f1a\u4fee\u6539\u4ee3\u7801\uff0c\u9488\u5bf9\u5f00\u53d1\u670d\u52a1\u5668\u8fd0\u884c\u5355\u5143\u6d4b\u8bd5\u6216 curl \u547d\u4ee4\uff0c\u770b\u5230\u901a\u8fc7\u7684\u7ed3\u679c\uff0c\u7136\u540e\u5c06\u529f\u80fd\u6807\u8bb0\u4e3a\u5df2\u5b8c\u6210\u3002\u4f46\u5f53\u7528\u6237\u901a\u8fc7\u6d4f\u89c8\u5668\u6d4b\u8bd5\u65f6\uff0c\u8be5\u529f\u80fd\u5b9e\u9645\u4e0a\u5e76\u4e0d\u8d77\u4f5c\u7528\u3002<\/p>\n<p>\u5355\u5143\u6d4b\u8bd5\u6210\u529f\u4e0e\u7aef\u5230\u7aef\u529f\u80fd\u4e4b\u95f4\u7684\u5dee\u8ddd\uff0c\u662f\u4eba\u7c7b\u5de5\u7a0b\u5e08\u901a\u8fc7\u5207\u6362\u4e0a\u4e0b\u6587\uff08\u8fd0\u884c\u5e94\u7528\u5e76\u5c1d\u8bd5\u4f7f\u7528\u5b83\uff09\u6765\u5904\u7406\u7684\u3002\u4e00\u4e2a\u6ca1\u6709\u660e\u786e\u6d4f\u89c8\u5668\u6d4b\u8bd5\u80fd\u529b\u7684\u667a\u80fd\u4f53\u65e0\u6cd5\u8fdb\u884c\u8fd9\u79cd\u5207\u6362\u3002\u5b83\u53ea\u80fd\u89c2\u5bdf\u5de5\u5177\u5141\u8bb8\u5b83\u89c2\u5bdf\u5230\u7684\u4e1c\u897f\uff0c\u5982\u679c\u8fd9\u4e9b\u5de5\u5177\u4e0d\u5305\u62ec\u6d4f\u89c8\u5668\u81ea\u52a8\u5316\uff0c\u5b83\u5c06\u4e00\u8d2f\u5730\u9057\u6f0f\u90a3\u4e9b\u4ec5\u5728\u771f\u5b9e\u7528\u6237\u6d41\u7a0b\u4e2d\u8868\u73b0\u51fa\u6765\u7684 Bug\u3002<\/p>\n<p>\u89e3\u51b3\u65b9\u6848\u662f\u8ba9\u667a\u80fd\u4f53\u8bbf\u95ee Puppeteer MCP \u670d\u52a1\u5668\uff0c\u8fd9\u662f\u4e00\u79cd\u6d4f\u89c8\u5668\u81ea\u52a8\u5316\u5de5\u5177\uff0c\u5141\u8bb8 Claude \u771f\u6b63\u5730\u5bfc\u822a\u5e94\u7528\u3001\u70b9\u51fb\u6309\u94ae\u3001\u586b\u5199\u8868\u5355\u5e76\u9a8c\u8bc1\u529f\u80fd\u662f\u5426\u7aef\u5230\u7aef\u8fd0\u884c\u3002\u6027\u80fd\u63d0\u5347\u662f\u5de8\u5927\u7684\u3002Bug\u4ec5\u4ece\u4ee3\u7801\u770b\u662f\u9690\u5f62\u7684\uff0c\u5728\u667a\u80fd\u4f53\u80fd\u770b\u5230\u7528\u6237\u6240\u89c1\u5185\u5bb9\u65f6\u53d8\u5f97\u663e\u800c\u6613\u89c1\u3002<\/p>\n<p>\u8fd9\u662f\u4e00\u4e2a\u666e\u9002\u539f\u5219\u7684\u5177\u4f53\u4f8b\u8bc1\uff1a<strong>\u667a\u80fd\u4f53\u5de5\u4f5c\u7684\u8d28\u91cf\u53d7\u9650\u4e8e\u5176\u53cd\u9988\u5faa\u73af\u7684\u8d28\u91cf<\/strong>\u3002\u5982\u679c\u4f60\u7684\u667a\u80fd\u4f53\u65e0\u6cd5\u5728\u5176\u6838\u5fc3\u9886\u57df\u89c2\u5bdf\u5176\u884c\u4e3a\u7684\u540e\u679c\uff0c\u5b83\u5c31\u4f1a\u53bb\u4f18\u5316\u90a3\u4e9b\u53ef\u80fd\u4e0e\u5b9e\u9645\u6b63\u786e\u6027\u65e0\u5173\u7684\u4ee3\u7406\u6307\u6807\uff08proxy metrics\uff09\u3002<\/p>\n<h4>\u542f\u52a8\u5e8f\u5217\uff1a\u5feb\u901f\u8fdb\u5165\u72b6\u6001<\/h4>\n<p>Anthropic \u7684 Harness \u4e2d\u6bcf\u4e00\u6b21\u7f16\u7801\u667a\u80fd\u4f53\u4f1a\u8bdd\u90fd\u4ee5\u6807\u51c6\u5316\u7684\u542f\u52a8\u5e8f\u5217\u5f00\u59cb\uff0c\u65e8\u5728\u5c3d\u53ef\u80fd\u5feb\u5730\u5f15\u5bfc\u667a\u80fd\u4f53\uff0c\u800c\u4e0d\u6d6a\u8d39\u4e0d\u5fc5\u8981\u7684 Token\u3002\u5e8f\u5217\u5982\u4e0b\uff1a<\/p>\n<p>1. \u00a0\u8fd0\u884c `pwd` \u4ee5\u786e\u8ba4\u5de5\u4f5c\u76ee\u5f55\u30022. \u00a0\u8bfb\u53d6\u8fdb\u5ea6\u6587\u4ef6\u548c git \u65e5\u5fd7\u4ee5\u4e86\u89e3\u8fd1\u671f\u5de5\u4f5c\u30023. \u00a0\u8bfb\u53d6\u529f\u80fd\u5217\u8868\u5e76\u9009\u62e9\u4f18\u5148\u7ea7\u6700\u9ad8\u7684\u672a\u5b8c\u6210\u529f\u80fd\u30024. \u00a0\u8fd0\u884c `init.sh` \u811a\u672c\u4ee5\u542f\u52a8\u5f00\u53d1\u73af\u5883\u30025. \u00a0\u8fd0\u884c\u57fa\u7840\u7aef\u5230\u7aef\u6d4b\u8bd5\u4ee5\u9a8c\u8bc1\u5e94\u7528\u5904\u4e8e\u5de5\u4f5c\u72b6\u6001\u3002<\/p>\n<p>\u53ea\u6709\u5728\u5b8c\u6210\u6240\u6709\u8fd9\u4e9b\u6b65\u9aa4\u540e\uff0c\u667a\u80fd\u4f53\u624d\u4f1a\u5f00\u59cb\u5904\u7406\u65b0\u529f\u80fd\u3002\u5982\u679c\u542f\u52a8\u6d4b\u8bd5\u663e\u793a\u5e94\u7528\u5df2\u635f\u574f\uff0c\u667a\u80fd\u4f53\u4f1a\u5728\u89e6\u78b0\u4efb\u4f55\u65b0\u4e1c\u897f\u4e4b\u524d\u5148\u4fee\u590d\u73b0\u6709\u7684\u635f\u574f\u3002\u8fd9\u9632\u6b62\u4e86\u95ee\u9898\u7684\u590d\u5408\u5316\u2014\u2014\u5373\u667a\u80fd\u4f53\u5728\u7834\u788e\u7684\u57fa\u7840\u4e0a\u5f00\u59cb\u65b0\u529f\u80fd\uff0c\u5bfc\u81f4\u5e95\u5c42\u95ee\u9898\u66f4\u96be\u88ab\u9694\u79bb\u548c\u4fee\u590d\u3002<\/p>\n<p>\u542f\u52a8\u5e8f\u5217\u8fd8\u4ee5\u7279\u5b9a\u65b9\u5f0f\u8282\u7701\u4e86 Token\u3002\u56e0\u4e3a `init.sh` \u811a\u672c\u8bb0\u5f55\u4e86\u542f\u52a8\u5f00\u53d1\u73af\u5883\u7684\u786e\u5207\u65b9\u6cd5\uff0c\u667a\u80fd\u4f53\u4e0d\u9700\u8981\u4ece\u5934\u6478\u7d22\u3002\u5728\u957f\u5468\u671f\u9879\u76ee\u4e2d\uff0c\u6bcf\u6b21\u4f1a\u8bdd\u5728\u73af\u5883\u8bbe\u7f6e\u4e0a\u8282\u7701\u7684 Token \u7d2f\u79ef\u8d77\u6765\u975e\u5e38\u663e\u8457\u3002<\/p>\n<div>\n<div class=\"hljs-cto\">\n<div class=\"hljs-cto\"><button class=\"copy_btn disable\" data-clipboard-target=\"#code_id_1\">\u590d\u5236<\/button><\/p>\n<div class=\"code-toolbar\">\n<pre class=\"has-pre-numbering language-plain\" tabindex=\"0\"><code class=\"language-plain\">[Assistant] I'll start by getting my bearings and understanding the current state of the project. [Tool Use] &lt;bash-pwd&gt; [Tool Use] &lt;read-claude-progress.txt&gt; [Tool Use] &lt;read-feature_list.json&gt; [Assistant] Let me check the git log to see recent work. [Tool Use] &lt;bash-gitlog--oneline-20&gt; [Assistant] Now let me check if there's an init.sh script to restart the servers. &lt;Startsthedevelopmentserver&gt; [Assistant] Excellent! Now let me navigate to the application and verify that some fundamental features are still working. &lt;Testsbasicfunctionality&gt;<\/code><\/pre>\n<ul id=\"code_id_1\" class=\"pre-numbering\">\n<li>1.<\/li>\n<\/ul>\n<div class=\"toolbar\"><\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<h3>OpenAI\u00a0\u7684 Harness \u5de5\u7a0b\uff08\u96f6\u884c\u4eba\u5de5\u4ee3\u7801\uff09<\/h3>\n<h4>\u5b9e\u9a8c\u80cc\u666f<\/h4>\n<p>2025 \u5e74 8 \u6708\u4e0b\u65ec\uff0cOpenAI \u7684 Codex \u56e2\u961f\u542f\u52a8\u4e86\u4e00\u4e2a git \u4ed3\u5e93\uff0c\u5e76\u8bbe\u7acb\u4e86\u4e00\u4e2a\u5355\u4e00\u7ea6\u675f\uff1a<strong>\u7981\u6b62\u4eba\u5de5\u7f16\u5199\u4ee3\u7801<\/strong>\u3002\u4ed3\u5e93\u4e2d\u7684\u6bcf\u4e00\u884c\u4ee3\u7801\uff0c\u5305\u62ec\u5e94\u7528\u903b\u8f91\u3001\u6d4b\u8bd5\u3001CI \u914d\u7f6e\u3001\u6587\u6863\u3001\u53ef\u89c2\u6d4b\u6027\u5de5\u5177\u548c\u5185\u90e8\u5f00\u53d1\u5de5\u5177\uff0c\u90fd\u5fc5\u987b\u7531 Codex \u667a\u80fd\u4f53\u7f16\u5199\u3002\u4eba\u7c7b\u8d1f\u8d23\u638c\u8235\uff0c\u667a\u80fd\u4f53\u8d1f\u8d23\u6267\u884c\u3002<\/p>\n<p>\u4e94\u4e2a\u6708\u540e\uff0c\u8be5\u4ed3\u5e93\u5728\u6240\u6709\u7c7b\u522b\u4e2d\u5305\u542b\u7ea6\u00a0<strong>100 \u4e07\u884c\u4ee3\u7801<\/strong>\u3002\u5927\u7ea6\u6709 1,500 \u4e2a\u62c9\u53d6\u8bf7\u6c42\uff08PR\uff09\u88ab\u5f00\u542f\u5e76\u5408\u5e76\u3002\u4e00\u652f\u7531\u4e09\u540d\u5de5\u7a0b\u5e08\u7ec4\u6210\u7684\u5c0f\u56e2\u961f\u9a71\u52a8\u4e86\u5176\u4e2d\u7684\u5927\u90e8\u5206\u5de5\u4f5c\uff0c\u5e73\u5747\u6bcf\u540d\u5de5\u7a0b\u5e08\u6bcf\u5929\u5904\u7406 3.5 \u4e2a PR\u3002\u968f\u7740\u56e2\u961f\u6269\u5927\u5230\u4e03\u540d\u5de5\u7a0b\u5e08\uff0c\u4eba\u5747\u541e\u5410\u91cf\u5b9e\u9645\u4e0a\u8fd8\u589e\u52a0\u4e86\u3002\u8be5\u4ea7\u54c1\u6bcf\u5929\u6709\u6570\u767e\u540d\u5185\u90e8\u7528\u6237\u548c\u5916\u90e8 Alpha \u6d4b\u8bd5\u4eba\u5458\u3002<\/p>\n<p>\u8fd9\u4e0d\u662f\u4e00\u4e2aDemo\u3002\u8fd9\u662f\u4e00\u4e2a\u5b8c\u5168\u901a\u8fc7\u667a\u80fd\u4f53\u751f\u6210\u7684\u4ee3\u7801\u6784\u5efa\u5e76\u4ea4\u4ed8\u7684\u771f\u5b9e\u5185\u90e8\u4ea7\u54c1\u3002\u8be5\u56e2\u961f\u5728 2026 \u5e74 2 \u6708\u64b0\u5199\u4e86\u63cf\u8ff0\u8fd9\u4e00\u7ecf\u5386\u7684\u6587\u7ae0\uff0c\u6838\u5fc3\u4fe1\u606f\u4e0e SWE-agent \u8bba\u6587\u4e00\u81f4\uff1a<strong>\u74f6\u9888\u4ece\u6765\u4e0d\u5728\u4e8e\u6a21\u578b\u80fd\u529b\uff0c\u800c\u59cb\u7ec8\u5728\u4e8e\u73af\u5883\u8bbe\u8ba1\u3002<\/strong><\/p>\n<h4>\u5de5\u7a0b\u5de5\u4f5c\u7684\u91cd\u65b0\u5b9a\u4e49<\/h4>\n<p>OpenAI \u7684 Harness \u5de5\u7a0b\u6587\u7ae0\u4e2d\u6700\u91cd\u8981\u7684\u89c2\u5bdf\u662f\u5173\u4e8e\u5de5\u7a0b\u5de5\u4f5c\u672c\u8eab\u662f\u5982\u4f55\u6f14\u53d8\u7684\u3002\u5f53\u4f60\u7684\u4e3b\u8981\u5de5\u4f5c\u4e0d\u518d\u662f\u7f16\u5199\u4ee3\u7801\u65f6\uff0c\u4f60\u8f6c\u800c\u5728\u505a\u4ec0\u4e48\uff1f<\/p>\n<p><strong>\u4f60\u5728\u8bbe\u8ba1\u73af\u5883\u3002\u4f60\u5728\u6307\u5b9a\u610f\u56fe\u3002\u4f60\u5728\u6784\u5efa\u53cd\u9988\u5faa\u73af\u3002<\/strong>\u4f60\u4e0d\u518d\u4e0d\u65ad\u5730\u95ee\u201c\u6211\u8be5\u5982\u4f55\u4fee\u590d\u8fd9\u4e2a Bug\uff1f\u201d\uff0c\u800c\u662f\u95ee\uff1a\u201c\u73af\u5883\u4e2d\u7f3a\u5c11\u4ec0\u4e48\u80fd\u529b\u5bfc\u81f4\u4e86\u8fd9\u4e2a Bug \u7684\u51fa\u73b0\uff1f\u201d<\/p>\n<p>\u5f53\u51fa\u73b0\u6545\u969c\u65f6\uff0c\u4fee\u590d\u65b9\u6cd5\u51e0\u4e4e\u4ece\u6765\u4e0d\u662f\u201c\u518d\u52aa\u529b\u4e00\u70b9\u201d\uff0c\u800c\u51e0\u4e4e\u603b\u662f\uff1a\u201c\u73af\u5883\u4e2d\u7f3a\u5c11\u6216\u8bef\u914d\u7f6e\u4e86\u4ec0\u4e48\u7ed3\u6784\u5316\u7ec4\u4ef6\uff0c\u5bfc\u81f4\u667a\u80fd\u4f53\u5728\u8fd9\u91cc\u5931\u8d25\uff1f\u201d\u8fd9\u662f\u5de5\u7a0b\u601d\u7ef4\u7684\u4e00\u4e2a\u6df1\u523b\u8f6c\u53d8\u3002<strong>\u4f60\u505c\u6b62\u8c03\u8bd5\u4ee3\u7801\uff0c\u8f6c\u800c\u5f00\u59cb\u8c03\u8bd5\u751f\u4ea7\u4ee3\u7801\u7684\u7cfb\u7edf\u3002<\/strong><\/p>\n<p>\u5de5\u7a0b\u56e2\u961f\u7684\u4e3b\u8981\u5de5\u4f5c\u53d8\u6210\u4e86<strong>\u8d4b\u80fd\u667a\u80fd\u4f53\u53bb\u5b8c\u6210\u6709\u7528\u7684\u5de5\u4f5c<\/strong>\uff0c\u800c\u4e0d\u662f\u7531\u4eba\u4eb2\u81ea\u53bb\u5b8c\u6210\u5de5\u4f5c\u3002<\/p>\n<p>\u5728\u5b9e\u8df5\u4e2d\uff0c\u8fd9\u610f\u5473\u7740\u5c06\u5b8f\u5927\u76ee\u6807\u5206\u89e3\u4e3a\u66f4\u5c0f\u7684\u6784\u5efa\u5757\uff0c\u6784\u5efa\u8ba9\u8fd9\u4e9b\u6784\u5efa\u5757\u53ef\u5b9e\u73b0\u7684\u5de5\u5177\u548c\u62bd\u8c61\uff0c\u5e76\u5c06\u5931\u8d25\u4f5c\u4e3a\u73af\u5883\u9700\u8981\u66f4\u597d\u652f\u6301\u4ec0\u4e48\u7684\u4fe1\u53f7\u3002\u4eba\u7c7b\u5de5\u7a0b\u5e08\u7684\u5de5\u4f5c\u662f\u201c\u6df1\u5ea6\u4f18\u5148\u201d\u7684\uff1a\u5f53\u667a\u80fd\u4f53\u5361\u4f4f\u65f6\uff0c\u4ed6\u4eec\u4e0d\u4f1a\u8bd5\u56fe\u81ea\u5df1\u5199\u4ee3\u7801\uff0c\u800c\u662f\u8be2\u95ee\u7f3a\u5c11\u4e86\u4ec0\u4e48\uff0c\u5c06\u5176\u6784\u5efa\u5230\u73af\u5883\u4e2d\uff0c\u7136\u540e\u8ba9\u667a\u80fd\u4f53\u91cd\u8bd5\u3002<\/p>\n<h4>\u4ed3\u5e93\u77e5\u8bc6\u4f5c\u4e3a\u8bb0\u5f55\u7cfb\u7edf<\/h4>\n<p>OpenAI \u7684 Harness \u4e2d\u6700\u91cd\u8981\u7684\u67b6\u6784\u51b3\u7b56\u4e4b\u4e00\uff0c\u662f<strong>\u5c06\u4ed3\u5e93\u672c\u8eab\u4f5c\u4e3a\u667a\u80fd\u4f53\u9700\u8981\u4e86\u89e3\u7684\u4e00\u5207\u4fe1\u606f\u7684\u6743\u5a01\u6765\u6e90<\/strong>\u3002\u8fd9\u4e00\u6d1e\u5bdf\u7b80\u5355\u5374\u6df1\u8fdc\uff1a\u4ece\u667a\u80fd\u4f53\u7684\u89c6\u89d2\u6765\u770b\uff0c\u4efb\u4f55\u5b83\u5728\u8fd0\u884c\u65f6\u65e0\u6cd5\u901a\u8fc7\u4e0a\u4e0b\u6587\u8bbf\u95ee\u7684\u4e1c\u897f\uff0c\u5b9e\u9645\u4e0a\u90fd\u4e0d\u5b58\u5728\u3002\u5b58\u5728\u4e8e Google Docs\u3001Slack \u5bf9\u8bdd\u6216\u4eba\u4eec\u5927\u8111\u4e2d\u7684\u77e5\u8bc6\uff0c\u5bf9\u7cfb\u7edf\u6765\u8bf4\u662f\u4e0d\u53ef\u89c1\u7684\u3002<\/p>\n<p>\u5728\u9879\u76ee\u65e9\u671f\uff0c\u56e2\u961f\u5c1d\u8bd5\u8fc7\u201c\u4e00\u4efd\u8d85\u5927\u7684 AGENTS.md\u201d\u65b9\u6848\u3002\u8fd9\u662f\u4e00\u4e2a\u5355\u4e00\u7684\u5927\u578b\u6307\u4ee4\u6587\u4ef6\uff0c\u5305\u542b\u667a\u80fd\u4f53\u9700\u8981\u4e86\u89e3\u7684\u5173\u4e8e\u9879\u76ee\u3001\u67b6\u6784\u3001\u7ea6\u5b9a\u548c\u7ea6\u675f\u7684\u4e00\u5207\u3002\u4e0d\u51fa\u6240\u6599\uff0c\u5b83\u5931\u8d25\u4e86\uff0c\u5176\u5931\u8d25\u6a21\u5f0f\u6709\u56db\u70b9\u503c\u5f97\u7406\u89e3\uff1a<\/p>\n<ol class=\"list-paddingleft-1\" start=\"1\" data-id=\"o4d4efab-kd1PFBY9\">\n<li data-id=\"ld70c578-YiQUQ3eE\"><strong>\u4e0a\u4e0b\u6587\u662f\u7a00\u7f3a\u8d44\u6e90<\/strong>\uff1a\u5de8\u578b\u6307\u4ee4\u6587\u4ef6\u6324\u5360\u4e86\u4efb\u52a1\u3001\u4ee3\u7801\u548c\u76f8\u5173\u6587\u6863\u7684\u7a7a\u95f4\u3002\u667a\u80fd\u4f53\u8981\u4e48\u9057\u6f0f\u5173\u952e\u7ea6\u675f\uff0c\u8981\u4e48\u5f00\u59cb\u9488\u5bf9\u9519\u8bef\u7684\u76ee\u6807\u8fdb\u884c\u4f18\u5316\u3002<\/li>\n<li data-id=\"ld70c578-dEleJeHA\"><strong>\u8fc7\u591a\u7684\u6307\u5bfc\u7b49\u4e8e\u6ca1\u6709\u6307\u5bfc<\/strong>\uff1a\u5f53\u6bcf\u4ef6\u4e8b\u90fd\u88ab\u6807\u8bb0\u4e3a\u91cd\u8981\u65f6\uff0c\u5c31\u6ca1\u6709\u4ec0\u4e48\u662f\u91cd\u8981\u7684\u4e86\u3002\u667a\u80fd\u4f53\u5f00\u59cb\u5728\u5c40\u90e8\u8fdb\u884c\u6a21\u5f0f\u5339\u914d\uff0c\u800c\u4e0d\u662f\u6709\u610f\u8bc6\u5730\u5bfc\u822a\u3002<\/li>\n<li data-id=\"ld70c578-5S4GYcjk\"><strong>\u8fc5\u901f\u8150\u70c2<\/strong>\uff1a\u968f\u7740\u4ee3\u7801\u5e93\u7684\u6f14\u8fdb\uff0c\u5355\u4f53\u624b\u518c\u4f1a\u53d8\u6210\u9648\u65e7\u89c4\u5219\u7684\u575f\u5893\u3002<\/li>\n<li data-id=\"ld70c578-ZECXcp1D\"><strong>\u96be\u4ee5\u9a8c\u8bc1<\/strong>\uff1a\u5355\u4e00\u7684\u5927\u5757\u5185\u5bb9\u4e0d\u5229\u4e8e\u8986\u76d6\u7387\u68c0\u67e5\u3001\u65b0\u9c9c\u5ea6\u8ffd\u8e2a\u6216\u4ea4\u53c9\u5f15\u7528\u3002\u504f\u5dee\u662f\u4e0d\u53ef\u907f\u514d\u7684\u3002<\/li>\n<\/ol>\n<p>\u89e3\u51b3\u65b9\u6848\u662f\u4e00\u4e2a<strong>\u7ed3\u6784\u5316\u7684\u00a0<\/strong><code><strong>docs\/<\/strong><\/code><strong>\u00a0\u76ee\u5f55<\/strong>\uff0c\u5c06\u5176\u89c6\u4e3a\u8bb0\u5f55\u7cfb\u7edf\uff0c\u5e76\u8f85\u4ee5\u4e00\u4e2a\u7b80\u77ed\u7684\u00a0<code>AGENTS.md<\/code>\u00a0\u6587\u4ef6\uff08\u7ea6 100 \u884c\uff09\u4f5c\u4e3a\u5730\u56fe\uff0c\u6307\u5411\u522b\u5904\u66f4\u6df1\u5c42\u6b21\u7684\u771f\u7406\u6765\u6e90\u3002\u8bbe\u8ba1\u6587\u6863\u88ab\u7f16\u76ee\u5e76\u7d22\u5f15\uff1b\u67b6\u6784\u6587\u6863\u63d0\u4f9b\u4e86\u9886\u57df\u548c\u5305\u5206\u5c42\u7684\u9876\u7ea7\u5730\u56fe\uff1b\u8ba1\u5212\u88ab\u89c6\u4e3a\u4e00\u7b49\u8d44\u4ea7\uff0c\u5176\u8fdb\u5ea6\u548c\u51b3\u7b56\u65e5\u5fd7\u90fd\u7b7e\u5165\u4ed3\u5e93\u3002<\/p>\n<p>\u8fd9\u5b9e\u73b0\u4e86\u56e2\u961f\u6240\u79f0\u7684\u201c<strong>\u6e10\u8fdb\u5f0f\u62ab\u9732<\/strong>\u201d\uff1a\u667a\u80fd\u4f53\u4ece\u4e00\u4e2a\u5fae\u5c0f\u3001\u7a33\u5b9a\u7684\u5165\u53e3\u70b9\u5f00\u59cb\uff0c\u5e76\u88ab\u6559\u5bfc\u63a5\u4e0b\u6765\u8be5\u53bb\u54ea\u91cc\u67e5\u627e\uff0c\u800c\u4e0d\u662f\u9884\u5148\u88ab\u6d77\u91cf\u4fe1\u606f\u6df9\u6ca1\u3002\u7ed3\u679c\u662f\uff0c\u667a\u80fd\u4f53\u53ef\u4ee5\u76f4\u63a5\u4ece\u4ed3\u5e93\u4e2d\u63a8\u7406\u51fa\u5b8c\u6574\u7684\u4e1a\u52a1\u9886\u57df\uff0c\u800c\u65e0\u9700\u8bbf\u95ee\u53ef\u80fd\u4e0d\u53ef\u7528\u6216\u5df2\u8fc7\u65f6\u7684\u5916\u90e8\u4e0a\u4e0b\u6587\u3002<\/p>\n<h4>\u5e94\u7528\u7684\u53ef\u8bfb\u6027\uff1a\u4f7f\u7cfb\u7edf\u5bf9\u667a\u80fd\u4f53\u53ef\u89c1<\/h4>\n<p>\u968f\u7740\u4ee3\u7801\u541e\u5410\u91cf\u7684\u589e\u52a0\uff0c\u74f6\u9888\u4ece\u201c\u751f\u6210\u201d\u8f6c\u79fb\u5230\u4e86\u201c\u9a8c\u8bc1\u201d\u3002\u56e2\u961f\u751f\u6210\u4ee3\u7801\u7684\u901f\u5ea6\u8d85\u8fc7\u4e86\u4eba\u7c7b QA\uff08\u8d28\u91cf\u4fdd\u8bc1\uff09\u80fd\u529b\u7684\u9a8c\u8bc1\u901f\u5ea6\u3002\u89e3\u51b3\u65b9\u6848\u662f<strong>\u8ba9\u5e94\u7528\u5bf9 Codex \u76f4\u63a5\u53ef\u8bfb<\/strong>\uff0c\u4ece\u800c\u8ba9\u66f4\u591a\u7684\u9a8c\u8bc1\u5de5\u4f5c\u53ef\u4ee5\u7531\u667a\u80fd\u4f53\u81ea\u884c\u5b8c\u6210\u3002<\/p>\n<p>\u8fd9\u6d89\u53ca\u51e0\u9879\u5177\u4f53\u7684\u6295\u5165\u3002\u4ed6\u4eec\u8ba9\u5e94\u7528\u53ef\u4ee5<strong>\u6309 git \u5de5\u4f5c\u6811 (worktree) \u542f\u52a8<\/strong>\uff0c\u56e0\u6b64 Codex \u53ef\u4ee5\u5728\u5b83\u6b63\u5728\u5904\u7406\u7684\u6bcf\u4e2a\u66f4\u6539\u4e2d\uff0c\u542f\u52a8\u5e76\u9a71\u52a8\u4e00\u4e2a\u9694\u79bb\u7684\u5e94\u7528\u5b9e\u4f8b\u3002\u4ed6\u4eec\u5c06\u00a0<strong>Chrome DevTools \u534f\u8bae<\/strong>\u63a5\u5165\u667a\u80fd\u4f53\u8fd0\u884c\u65f6\uff0c\u5e76\u521b\u5efa\u4e86\u5904\u7406 DOM \u5feb\u7167\u3001\u5c4f\u5e55\u622a\u56fe\u548c\u6d4f\u89c8\u5668\u5bfc\u822a\u7684\u5de5\u5177\u3002\u8fd9\u4f7f\u5f97 Codex \u80fd\u591f\u76f4\u63a5\u590d\u73b0 Bug\u3001\u9a8c\u8bc1\u4fee\u590d\u5e76\u63a8\u7406 UI \u884c\u4e3a\uff0c\u800c\u65e0\u9700\u4eba\u7c7b\u4e0e\u5e94\u7528\u4ea4\u4e92\u3002<\/p>\n<p>\u4ed6\u4eec\u6784\u5efa\u4e86\u4e00\u4e2a\u5b8c\u6574\u7684<strong>\u672c\u5730\u53ef\u89c2\u6d4b\u6027\u6808<\/strong>\uff1a\u901a\u8fc7 LogQL\u3001PromQL \u548c TraceQL \u5411 Codex \u5f00\u653e\u65e5\u5fd7\u3001\u6307\u6807\u548c\u8ffd\u8e2a\u3002\u6bcf\u4e2a\u667a\u80fd\u4f53\u4efb\u52a1\u90fd\u5728\u4e00\u4e2a\u5b8c\u5168\u9694\u79bb\u7684\u5e94\u7528\u7248\u672c\u4e0a\u8fd0\u884c\uff0c\u5e76\u62e5\u6709\u81ea\u5df1\u7684\u89c2\u6d4b\u6570\u636e\uff0c\u4efb\u52a1\u5b8c\u6210\u540e\u5373\u9500\u6bc1\u3002\u8fd9\u610f\u5473\u7740\u667a\u80fd\u4f53\u53ef\u4ee5\u4f7f\u7528\u771f\u5b9e\u7684\u53ef\u89c2\u6d4b\u6027\u5de5\u5177\uff08\u4e0e\u4eba\u7c7b\u5de5\u7a0b\u5e08\u4f7f\u7528\u7684\u5de5\u5177\u76f8\u540c\uff09\u6765\u8c03\u8bd5\u751f\u4ea7\u7ea7\u95ee\u9898\uff0c\u800c\u4e0d\u662f\u4ec5\u51ed\u4ee3\u7801\u63a8\u6d4b\u884c\u4e3a\u3002<\/p>\n<p>\u8fd9\u91cc\u7684\u539f\u5219\u4e0e SWE-agent \u8bba\u6587\u6240\u8bc1\u660e\u7684\u4e00\u81f4\uff1a<strong>\u667a\u80fd\u4f53\u5de5\u4f5c\u7684\u8d28\u91cf\u53d7\u9650\u4e8e\u5176\u53cd\u9988\u5faa\u73af\u7684\u8d28\u91cf\u3002<\/strong>\u00a0\u5982\u679c\u667a\u80fd\u4f53\u80fd\u770b\u5230\u7528\u6237\u6240\u89c1\uff0c\u5e76\u80fd\u89c2\u5bdf\u5230\u4eba\u7c7b\u5de5\u7a0b\u5e08\u6240\u89c2\u5bdf\u7684\u76f8\u540c\u6307\u6807\u548c\u65e5\u5fd7\uff0c\u5b83\u5c31\u80fd\u6355\u6349\u5e76\u4fee\u590d\u6bd4\u4ec5\u64cd\u4f5c\u4ee3\u7801\u7684\u667a\u80fd\u4f53\u591a\u5f97\u591a\u7684\u95ee\u9898\u3002<\/p>\n<h4>\u5728\u4e0d\u8fdb\u884c\u5fae\u89c2\u7ba1\u7406\u7684\u60c5\u51b5\u4e0b\u5f3a\u5316\u67b6\u6784<\/h4>\n<p>\u5728\u5168\u667a\u80fd\u4f53\u751f\u6210\u7684\u4ee3\u7801\u5e93\u4e2d\uff0c\u6700\u5927\u7684\u6311\u6218\u4e4b\u4e00\u662f\u957f\u671f\u4fdd\u6301\u67b6\u6784\u7684\u8fde\u8d2f\u6027\u3002Codex \u4f1a\u590d\u5236\u4ed3\u5e93\u4e2d\u5df2\u6709\u7684\u6a21\u5f0f\uff0c\u5305\u62ec\u90a3\u4e9b\u4e0d\u5747\u8861\u6216\u6b21\u4f18\u7684\u6a21\u5f0f\u3002\u4e45\u800c\u4e45\u4e4b\uff0c\u8fd9\u4f1a\u5bfc\u81f4\u504f\u5dee\uff1a\u574f\u6a21\u5f0f\u8513\u5ef6\uff0c\u4e0d\u4e00\u81f4\u6027\u7d2f\u79ef\uff0c\u4ee3\u7801\u5e93\u53d8\u5f97\u8d8a\u6765\u8d8a\u96be\u8ba9\u672a\u6765\u7684\u667a\u80fd\u4f53\u6b63\u786e\u5bfc\u822a\u3002<\/p>\n<p>OpenAI \u7684\u89e3\u51b3\u65b9\u6848\u662f<strong>\u673a\u68b0\u5316\u5730\u5f3a\u5236\u6267\u884c\u4e0d\u53d8\u6027\u00a0(Invariants)<\/strong>\uff0c\u800c\u4e0d\u662f\u901a\u8fc7\u4eba\u5de5\u4ee3\u7801\u5ba1\u67e5\u3002\u5e94\u7528\u56f4\u7ed5\u4e00\u4e2a\u4e25\u82db\u7684\u67b6\u6784\u6a21\u578b\u6784\u5efa\uff1a\u6bcf\u4e2a\u4e1a\u52a1\u9886\u57df\u88ab\u5212\u5206\u4e3a\u4e00\u7ec4\u56fa\u5b9a\u7684\u5c42\u7ea7\uff0c\u5177\u6709\u4e25\u683c\u9a8c\u8bc1\u7684\u4f9d\u8d56\u65b9\u5411\u548c\u6709\u9650\u7684\u5141\u8bb8\u8fb9\u7f18\u3002\u8fd9\u4e9b\u7ea6\u675f\u7531\u81ea\u5b9a\u4e49\u7684\u00a0<strong>Linter<\/strong>\uff08\u81ea\u7136\u4e5f\u662f\u7531 Codex \u7f16\u5199\u7684\uff09\u548c\u7ed3\u6784\u5316\u6d4b\u8bd5\u6765\u5f3a\u5236\u6267\u884c\u3002<\/p>\n<p>\u6838\u5fc3\u6d1e\u5bdf\u662f\u5728\u5f3a\u5316\u8fb9\u754c\u7684\u540c\u65f6\uff0c\u5141\u8bb8\u5728\u8fb9\u754c\u5185\u62e5\u6709\u5de8\u5927\u7684\u81ea\u7531\u3002Linter \u68c0\u67e5\u4ee3\u7801\u662f\u5426\u5728\u5c42\u7ea7\u7ed3\u6784\u4e2d\u6309\u6b63\u786e\u65b9\u5411\u6d41\u52a8\uff0c\u4f46\u5b83\u4eec\u4e0d\u5e72\u9884\u7279\u5b9a\u529f\u80fd\u5728\u8fd9\u4e9b\u8fb9\u754c\u5185\u662f\u5982\u4f55\u5b9e\u73b0\u7684\u3002\u8fd9\u4e0e\u8ba9\u5e73\u53f0\u56e2\u961f\u5728\u89c4\u6a21\u5316\u65f6\u53d1\u6325\u4f5c\u7528\u7684\u539f\u5219\u76f8\u540c\uff1a<strong>\u5f3a\u5316\u57fa\u5ea7\uff0c\u5141\u8bb8\u5176\u4e0a\u7684\u81ea\u4e3b\u6743\u3002<\/strong><\/p>\n<p>\u8fd9\u4e9b Linter \u662f\u4e13\u95e8\u4e3a\u751f\u6210\u5bf9\u667a\u80fd\u4f53\u6709\u7528\u7684\u9519\u8bef\u4fe1\u606f\u800c\u7f16\u5199\u7684\u3002\u5f53 Linter \u6355\u6349\u5230\u8fdd\u89c4\u65f6\uff0c\u9519\u8bef\u4fe1\u606f\u4e2d\u5305\u542b\u4e86\u683c\u5f0f\u5316\u7684<strong>\u4fee\u590d\u6307\u4ee4<\/strong>\uff0c\u4ee5\u4fbf\u6ce8\u5165\u667a\u80fd\u4f53\u7684\u4e0a\u4e0b\u6587\u3002\u8fd9\u5173\u95ed\u4e86\u95ed\u73af\uff1a\u88ab\u8fdd\u53cd\u7684\u7ea6\u675f\u3001\u88ab\u8fdd\u53cd\u7684\u89c4\u5219\u4ee5\u53ca\u4fee\u590d\u6b65\u9aa4\uff0c\u90fd\u901a\u8fc7\u4e00\u6761\u5355\u4e00\u7684\u3001\u53ef\u64cd\u4f5c\u7684\u53cd\u9988\u4fe1\u606f\u9001\u8fbe\u3002<\/p>\n<p>\u4ed6\u4eec\u8fd8\u5c06\u6240\u8c13\u7684\u201c<strong>\u9ec4\u91d1\u539f\u5219 (Golden Principles)<\/strong>\u201d\u76f4\u63a5\u7f16\u7801\u8fdb\u4ed3\u5e93\uff1a\u8fd9\u4e9b\u662f\u6b66\u65ad\u7684\u3001\u673a\u68b0\u7684\u89c4\u5219\uff0c\u65e8\u5728\u4fdd\u6301\u4ee3\u7801\u5e93\u5bf9\u672a\u6765\u667a\u80fd\u4f53\u8fd0\u884c\u7684\u53ef\u8bfb\u6027\u548c\u4e00\u81f4\u6027\u3002\u4f8b\u5982\uff1a\u4f18\u5148\u4f7f\u7528\u5171\u4eab\u5de5\u5177\u5305\u800c\u975e\u624b\u5199\u7684\u52a9\u624b\u51fd\u6570\uff1b\u5728\u8fb9\u754c\u5904\u9a8c\u8bc1\u6570\u636e\u5f62\u72b6\u3002\u8fd9\u4e9b\u539f\u5219\u7531\u5b9a\u671f\u8fd0\u884c\u7684\u6e05\u7406\u540e\u53f0\u4efb\u52a1\u5f3a\u5236\u6267\u884c\uff0c\u8fd9\u4e9b\u4efb\u52a1\u4f1a\u626b\u63cf\u504f\u5dee\u3001\u66f4\u65b0\u8d28\u91cf\u7b49\u7ea7\uff0c\u5e76\u5f00\u542f\u9488\u5bf9\u6027\u7684\u91cd\u6784 PR\u3002\u5176\u4e2d\u5927\u591a\u6570 PR \u53ef\u4ee5\u5728\u4e00\u5206\u949f\u5185\u5b8c\u6210 Review \u5e76\u81ea\u52a8\u5408\u5e76\u3002<\/p>\n<h4>\u541e\u5410\u91cf\u6539\u53d8\u4e86\u5408\u5e76\u54f2\u5b66<\/h4>\n<p>\u5f53\u667a\u80fd\u4f53\u7684\u541e\u5410\u91cf\u5927\u5e45\u8d85\u8fc7\u4eba\u7c7b\u7684\u6ce8\u610f\u529b\u5bb9\u91cf\u65f6\uff0c\u4f20\u7edf\u7684\u5de5\u7a0b\u89c4\u8303\u5c31\u4f1a\u4ea7\u751f\u53cd\u4f5c\u7528\u3002\u7b49\u5f85 Review \u7684 PR \u4f1a\u963b\u585e\u667a\u80fd\u4f53\u7684\u5de5\u4f5c\uff1b\u88ab\u9010\u4e00\u8c03\u67e5\u7684\u6d4b\u8bd5\u95ea\u70c1\uff08Test flakes\uff09\u4f1a\u6d88\u8017\u672c\u53ef\u4ee5\u6295\u5165\u5230\u66f4\u9ad8\u6760\u6746\u4efb\u52a1\u4e0a\u7684\u4eba\u7c7b\u6ce8\u610f\u529b\u3002<\/p>\n<p>OpenAI \u56e2\u961f\u505a\u51fa\u4e86\u4e00\u4e2a\u6df1\u601d\u719f\u8651\u7684\u51b3\u5b9a\uff1a<strong>\u4ee5\u6700\u5c0f\u5316\u7684\u963b\u585e\u5f0f\u5408\u5e76\u95e8\u7981\u6765\u8fd0\u884c\u3002<\/strong>\u00a0PR \u4fdd\u6301\u77ed\u751f\u547d\u5468\u671f\u3002\u6d4b\u8bd5\u95ea\u70c1\u901a\u8fc7\u540e\u7eed\u8fd0\u884c\u6765\u5904\u7406\uff0c\u800c\u4e0d\u662f\u65e0\u9650\u671f\u963b\u585e\u8fdb\u5ea6\u3002<strong>\u5f53\u667a\u80fd\u4f53\u541e\u5410\u91cf\u8fdc\u8d85\u4eba\u7c7b\u6ce8\u610f\u529b\u65f6\uff0c\u7ea0\u9519\u662f\u5ec9\u4ef7\u7684\uff0c\u800c\u7b49\u5f85\u662f\u6602\u8d35\u7684\u3002<\/strong>\u00a0\u8fd9\u79cd\u6743\u8861\u5728\u4f4e\u541e\u5410\u91cf\u73af\u5883\u4e0b\u770b\u8d77\u6765\u662f\u4e0d\u8d1f\u8d23\u4efb\u7684\uff0c\u4f46\u5728\u9ad8\u541e\u5410\u91cf\u73af\u5883\u4e0b\u5374\u662f\u663e\u800c\u6613\u89c1\u7684\u3002<\/p>\n<p>\u5bf9\u4e8e\u6b63\u5728\u8f6c\u5411\u667a\u80fd\u4f53\u9a71\u52a8\u5f00\u53d1\u7684\u56e2\u961f\u6765\u8bf4\uff0c\u8fd9\u662f\u4e00\u4e2a\u771f\u6b63\u91cd\u8981\u7684\u6d1e\u5bdf\u3002\u5f53\u4eba\u7c7b\u5de5\u7a0b\u5e08\u7f16\u5199\u6bcf\u4e00\u884c\u4ee3\u7801\u65f6\u5408\u7406\u7684\u5408\u5e76\u54f2\u5b66\uff0c\u5728\u667a\u80fd\u4f53\u6bcf\u4eba\u6bcf\u5929\u751f\u6210 3.5 \u4e2a PR \u65f6\uff0c\u5e76\u4e0d\u4e00\u5b9a\u81ea\u52a8\u5408\u7406\u3002\u74f6\u9888\u8f6c\u79fb\u4e86\uff0c\u6d41\u7a0b\u4e5f\u9700\u8981\u968f\u4e4b\u8f6c\u53d8\u3002<\/p>\n<h3>Awesome Agent Harness\u5206\u7c7b\u6cd5<\/h3>\n<h4>\u63cf\u7ed8\u751f\u6001\u7cfb\u7edf\u7248\u56fe<\/h4>\n<p>\u7531 GitHub \u4e0a\u7684 AutoJunjie \u9879\u76ee\u7ef4\u62a4\u7684 Awesome Agent Harness \u4ee3\u7801\u4ed3\u5e93\uff0c\u8bd5\u56fe\u63cf\u7ed8\u65b0\u5174\u7684harness\u5de5\u7a0b\u5de5\u5177\u7684\u751f\u6001\u7cfb\u7edf\u7248\u56fe\u3002\u5728\u6df1\u5165\u63a2\u8ba8\u8be5\u5206\u7c7b\u6cd5\u4e4b\u524d\uff0c\u5176\u6838\u5fc3\u8bba\u70b9\u503c\u5f97\u660e\u786e\u9648\u8ff0\uff1a\u4eba\u5de5\u667a\u80fd\u7f16\u5199\u4ee3\u7801\u7684\u80fd\u529b\u5b9e\u9645\u4e0a\u5df2\u6210\u4e3a\u4e00\u79cd\u5546\u54c1\uff08commodity\uff09\u3002\u57fa\u7840\u6a21\u578b\u80fd\u591f\u751f\u6210\u5177\u6709\u529f\u80fd\u6027\u7684\u4ee3\u7801\u3002\u8fd9\u5df2\u4e0d\u518d\u662f\u4ea7\u751f\u5dee\u5f02\u5316\u7684\u80fd\u529b\u3002\u771f\u6b63\u4ea7\u751f\u5dee\u5f02\u5316\u7684\u80fd\u529b\u5728\u4e8e\u534f\u8c03\u4e0e\u73af\u5883\u8bbe\u8ba1\u3002<\/p>\n<p>\u8be5\u4ee3\u7801\u4ed3\u5e93\u5c06\u4e00\u4e2a\u4e25\u8083\u7684\u667a\u80fd\u4f53Harness\u751f\u6001\u7cfb\u7edf\u6240\u9700\u7684\u5b8c\u6574\u6280\u672f\u6808\u8fdb\u884c\u4e86\u76ee\u5f55\u5316\u6574\u7406\uff0c\u5206\u4e3a\u4e03\u4e2a\u622a\u7136\u4e0d\u540c\u7684\u5c42\u7ea7\u3002\u7406\u89e3\u8fd9\u4e9b\u5c42\u7ea7\u6709\u52a9\u4e8e\u89e3\u91ca\u4e3a\u4ec0\u4e48\u201c\u6784\u5efa\u4e00\u4e2a AI \u7f16\u7801\u52a9\u624b\u201d\u5b9e\u9645\u4e0a\u662f\u8bb8\u591a\u4e2a\u72ec\u7acb\u7684\u5de5\u7a0b\u95ee\u9898\uff0c\u800c\u4e0d\u662f\u5355\u4e00\u7684\u95ee\u9898\u3002<\/p>\n<h4>\u7b2c\u4e00\u5c42\uff1a\u4eba\u7c7b\u76d1\u7763<\/h4>\n<p>\u4f4d\u4e8e\u6700\u9876\u5c42\u7684\u662f\u4eba\u7c7b\u76d1\u7763\u5c42\uff0c\u4eba\u7c7b\u5728\u8fd9\u91cc\u6279\u51c6\u63d0\u6848\u3001\u5ba1\u67e5\u62c9\u53d6\u8bf7\u6c42\uff08pull requests\uff09\u5e76\u8bbe\u5b9a\u4f18\u5148\u7ea7\u3002\u8fd9\u4e0d\u662f\u4f20\u7edf\u610f\u4e49\u4e0a\u7684\u6280\u672f\u5c42\u3002\u5b83\u662f\u4eba\u7c7b\u5224\u65ad\u4e0e\u667a\u80fd\u4f53\u6267\u884c\u4e4b\u95f4\u7684\u63a5\u53e3\u3002\u8fd9\u91cc\u7684\u5173\u952e\u8bbe\u8ba1\u539f\u5219\u662f\uff0c\u5de5\u7a0b\u5e08\u5e94\u8be5\u662f\u5728\u8bbe\u8ba1\u73af\u5883\u548c\u5ba1\u67e5\u7ed3\u679c\uff0c\u800c\u4e0d\u662f\u76f4\u63a5\u7f16\u5199\u4ee3\u7801\u3002\u4ed6\u4eec\u7684\u6760\u6746\u4f5c\u7528\uff08\u5f71\u54cd\u529b\uff09\u6765\u81ea\u4e8e\u5f15\u5bfc\uff08steering\uff09\uff0c\u800c\u4e0d\u662f\u6267\u884c\u3002<\/p>\n<h4>\u7b2c\u4e8c\u5c42\uff1a\u89c4\u5212\u4e0e\u9700\u6c42\uff08\u89c4\u7ea6\u5de5\u5177\uff09<\/h4>\n<p>\u8fd9\u4e00\u5c42\u5c06\u4eba\u7c7b\u7684\u60f3\u6cd5\u8f6c\u5316\u4e3a\u7ed3\u6784\u5316\u7684\u89c4\u7ea6\u548c\u4efb\u52a1\u6709\u5411\u65e0\u73af\u56fe\uff08DAG\uff09\uff0c\u4ee5\u4fbf\u667a\u80fd\u4f53\u80fd\u591f\u53ef\u9760\u5730\u4f7f\u7528\u3002\u5176\u6f5c\u5728\u7684\u6df1\u523b\u89c1\u89e3\u662f\uff0c\u667a\u80fd\u4f53\u662f\u76f2\u76ee\u6267\u884c\u7684\u3002\u5982\u679c\u89c4\u7ea6\u6a21\u7cca\u6216\u5b58\u5728\u6b67\u4e49\uff0c\u667a\u80fd\u4f53\u5c06\u751f\u6210\u6ee1\u8db3\u5176\u5bf9\u8be5\u89c4\u7ea6\u7406\u89e3\u7684\u5185\u5bb9\uff0c\u800c\u8fd9\u53ef\u80fd\u5e76\u975e\u4eba\u7c7b\u7684\u672c\u610f\u3002\u89c4\u7ea6\u5de5\u5177\u5728\u7f16\u5199\u4efb\u4f55\u4ee3\u7801\u4e4b\u524d\uff0c\u5728\u9700\u6c42\u9636\u6bb5\u5c31\u5f3a\u5236\u8981\u6c42\u7cbe\u786e\u6027\u3002<\/p>\n<p>\u8be5\u9886\u57df\u7684\u4e00\u4e2a\u9879\u76ee Chorus\uff0c\u8bd5\u56fe\u89e3\u51b3\u4ee3\u7801\u4ed3\u5e93\u4e2d\u6240\u79f0\u7684\u201c\u53cd\u5411\u5bf9\u8bdd\u9e3f\u6c9f\uff08reversed conversation gap\uff09\u201d\u3002Chorus \u5e76\u6ca1\u6709\u8ba9\u81ea\u7136\u7f3a\u4e4f\u667a\u80fd\u4f53\u6240\u9700\u7cbe\u786e\u5ea6\u7684\u4eba\u7c7b\u53bb\u7f16\u5199\u8be6\u7ec6\u7684\u89c4\u7ea6\u5de5\u5355\uff08\u8fd9\u662f\u4e00\u4e2a\u4e3b\u8981\u7684\u5931\u8d25\u70b9\uff09\uff0c\u800c\u662f\u8ba9 AI \u63d0\u51fa\u4efb\u52a1 DAG \u5e76\u8be6\u7ec6\u9610\u8ff0\u9700\u6c42\uff0c\u4eba\u7c7b\u5728\u6267\u884c\u5f00\u59cb\u524d\u4ec5\u626e\u6f14\u4e25\u683c\u7684\u9a8c\u8bc1\u548c\u6279\u51c6\u89d2\u8272\u3002\u5728\u6839\u636e\u90e8\u5206\u610f\u56fe\u751f\u6210\u5b8c\u6574\u89c4\u7ea6\u65b9\u9762\uff0cAI \u6bd4\u4eba\u7c7b\u4ece\u96f6\u5f00\u59cb\u7f16\u5199\u8981\u64c5\u957f\u5f97\u591a\u3002<\/p>\n<h4>\u7b2c\u4e09\u5c42\uff1a\u5168\u751f\u547d\u5468\u671f\u5e73\u53f0<\/h4>\n<p>\u8fd9\u4e9b\u5de5\u5177\u7ba1\u7406\u4ece\u521d\u59cb\u9700\u6c42\u5230\u4ea4\u4ed8\u7684\u7aef\u5230\u7aef\u6d41\u7a0b\uff0c\u5c06 AI \u63d0\u6848\u4e0e\u4eba\u7c7b\u9a8c\u8bc1\u5173\u5361\u4ee5\u53ca\u5b50\u667a\u80fd\u4f53\u7f16\u6392\u6574\u5408\u5728\u4e00\u8d77\u3002\u5b83\u4eec\u662f\u89c4\u7ea6\u5c42\u548c\u6267\u884c\u5c42\u4e4b\u95f4\u7684\u7c98\u5408\u5242\uff0c\u5904\u7406\u8d2f\u7a7f\u6574\u4e2a\u5f00\u53d1\u751f\u547d\u5468\u671f\u7684\u72b6\u6001\u7ba1\u7406\u3002<\/p>\n<h4>\u7b2c\u56db\u5c42\uff1a\u4efb\u52a1\u8fd0\u884c\u5668<\/h4>\n<p>\u4efb\u52a1\u8fd0\u884c\u5668\u5f25\u5408\u4e86\u95ee\u9898\u8ddf\u8e2a\u5668\uff08\u5982 GitHub Issues\u3001Linear\uff09\u4e0e\u7f16\u7801\u667a\u80fd\u4f53\u4e4b\u95f4\u7684\u9e3f\u6c9f\u3002\u6d41\u7a0b\u5982\u4e0b\uff1a\u4eba\u7c7b\u6216\u4ea7\u54c1\u7ecf\u7406\uff08PM\uff09\u667a\u80fd\u4f53\u521b\u5efa\u4e00\u4e2a\u95ee\u9898\uff0c\u4efb\u52a1\u8fd0\u884c\u5668\u751f\u6210\u4e00\u4e2a\u5de5\u4f5c\u533a\uff0c\u667a\u80fd\u4f53\u4ea4\u4ed8\u4e00\u4e2aPR\uff0c\u7136\u540e\u4eba\u7c7b\u8fdb\u884c\u5ba1\u67e5\u3002\u6b64\u7c7b\u522b\u7684\u5de5\u5177\u5305\u62ec\u4e0d\u65ad\u8f6e\u8be2\u4efb\u52a1\u961f\u5217\u3001\u51b3\u5b9a\u4f55\u65f6\u751f\u6210\u667a\u80fd\u4f53\u5e76\u4ea4\u4ed8\u5df2\u5b8c\u6210\u5de5\u4f5c\u7684\u7cfb\u7edf\uff0c\u800c\u65e0\u9700\u4eba\u7c7b\u53c2\u4e0e\u6267\u884c\u5faa\u73af\u3002<\/p>\n<h4>\u7b2c\u4e94\u5c42\uff1a\u667a\u80fd\u4f53\u7f16\u6392\u5668<\/h4>\n<p>\u7f16\u6392\u5668\u901a\u8fc7\u5b9e\u73b0\u591a\u4e2a\u667a\u80fd\u4f53\u7684\u5e76\u884c\u6267\u884c\uff0c\u540c\u65f6\u5c06\u5b83\u4eec\u7684\u5de5\u4f5c\u9694\u79bb\u5728\u72ec\u7acb\u7684 git \u5de5\u4f5c\u6811\uff08worktrees\uff09\u4e2d\uff0c\u4ece\u800c\u89e3\u51b3\u541e\u5410\u91cf\u95ee\u9898\u3002\u8fd9\u81f3\u5173\u91cd\u8981\uff0c\u56e0\u4e3a\u5982\u679c\u5171\u4eab\u540c\u4e00\u4e2a\u5de5\u4f5c\u533a\uff0c\u5728\u540c\u4e00\u4ee3\u7801\u5e93\u4e0a\u5e76\u884c\u5de5\u4f5c\u7684\u667a\u80fd\u4f53\u4f1a\u76f8\u4e92\u51b2\u7a81\u3002Git \u5de5\u4f5c\u6811\u9694\u79bb\u4e3a\u6bcf\u4e2a\u667a\u80fd\u4f53\u63d0\u4f9b\u4e86\u81ea\u5df1\u7684\u6c99\u7bb1\uff0c\u5141\u8bb8\u8bb8\u591a\u667a\u80fd\u4f53\u540c\u65f6\u5de5\u4f5c\u800c\u4e0d\u4f1a\u4e92\u76f8\u8e29\u8e0f\uff08\u5e72\u6270\uff09\u3002<\/p>\n<p>\u8bf8\u5982 Vibe Kanban\u3001Emdash \u548c Composio \u4e4b\u7c7b\u7684\u5de5\u5177\u5b9e\u73b0\u4e86\u8fd9\u79cd\u6a21\u5f0f\u3002\u6bcf\u4e2a\u667a\u80fd\u4f53\u4efb\u52a1\u90fd\u4f1a\u83b7\u5f97\u81ea\u5df1\u7684 git \u5de5\u4f5c\u6811\u3002\u66f4\u6539\u5728\u5408\u5e76\u4e4b\u524d\u4f1a\u5728\u9694\u79bb\u72b6\u6001\u4e0b\u8fdb\u884c\u9a8c\u8bc1\u3002\u6301\u7eed\u96c6\u6210\uff08CI\uff09\u53cd\u9988\u3001\u5408\u5e76\u51b2\u7a81\u4ee5\u53ca\u667a\u80fd\u4f53\u4e4b\u95f4\u7684\u534f\u8c03\u90fd\u7531\u7f16\u6392\u5c42\u5904\u7406\uff0c\u65e0\u9700\u4eba\u5de5\u5e72\u9884\u3002<\/p>\n<h4>\u7b2c\u516d\u5c42\uff1a\u667a\u80fd\u4f53Harness\u6846\u67b6\u4e0e\u8fd0\u884c\u65f6<\/h4>\n<p>\u6846\u67b6\uff08Frameworks\uff09\u4e3a\u6784\u5efa\u81ea\u5b9a\u4e49\u73af\u5883\u63d0\u4f9b\u4e86\u53ef\u7ec4\u5408\u7684\u57fa\u7840\uff1a\u6e10\u8fdb\u5f0f\u62ab\u9732\u673a\u5236\u3001\u5b50\u667a\u80fd\u4f53\u751f\u6210\u3001\u7ed3\u6784\u5316\u7684\u4e0a\u4e0b\u6587\u4ea4\u4ed8\u3002\u8fd0\u884c\u65f6\uff08Runtimes\uff09\u63d0\u4f9b\u6301\u4e45\u7684\u57fa\u7840\u8bbe\u65bd\uff1a\u957f\u671f\u8fd0\u884c\u7684\u5185\u5b58\u3001\u8ba1\u5212\uff08\u5b9a\u65f6\uff09\u6267\u884c\u3001\u4f1a\u8bdd\u4e4b\u95f4\u7684\u591a\u6e20\u9053\u901a\u4fe1\u3002<\/p>\n<p>\u6846\u67b6\u4e0e\u8fd0\u884c\u65f6\u4e4b\u95f4\u7684\u533a\u522b\u975e\u5e38\u91cd\u8981\u3002\u6846\u67b6\u662f\u4f60\u6784\u5efa\u7684\u57fa\u7840\u3002\u8fd0\u884c\u65f6\u662f\u6301\u7eed\u8fd0\u884c\u7684\u5b9e\u4f53\u3002Claude Agent SDK \u4e3b\u8981\u662f\u4e00\u4e2a\u6846\u67b6\u3002\u4e00\u4e2a\u6309 cron \u8ba1\u5212\u8fd0\u884c\u667a\u80fd\u4f53\u3001\u5728\u4f1a\u8bdd\u4e4b\u95f4\u4fdd\u6301\u6301\u4e45\u5185\u5b58\u3001\u5e76\u5904\u7406\u667a\u80fd\u4f53\u5b9e\u4f8b\u4e4b\u95f4\u591a\u6e20\u9053\u534f\u8c03\u7684\u7cfb\u7edf\uff0c\u5219\u662f\u4e00\u4e2a\u8fd0\u884c\u65f6\u3002\u5bf9\u4e8e\u4e25\u8083\u7684\u3001\u957f\u671f\u8fd0\u884c\u7684\u667a\u80fd\u4f53\u5316\u5de5\u4f5c\u6765\u8bf4\uff0c\u4e24\u8005\u90fd\u662f\u5fc5\u4e0d\u53ef\u5c11\u7684\u3002<\/p>\n<h4>\u7b2c\u4e03\u5c42\uff1a\u7f16\u7801\u667a\u80fd\u4f53<\/h4>\n<p>\u4f4d\u4e8e\u6700\u5e95\u5c42\u7684\u662f\u6267\u884c\u5c42\uff1aClaude Code\u3001Codex \u4ee5\u53ca\u7c7b\u4f3c\u7f16\u5199\u3001\u6d4b\u8bd5\u548c\u8c03\u8bd5\u4ee3\u7801\u7684\u7cfb\u7edf\u3002\u8be5\u4ee3\u7801\u4ed3\u5e93\u7684\u6838\u5fc3\u6d1e\u5bdf\u662f\uff0c\u8fd9\u4e00\u5c42\u5df2\u6210\u4e3a\u4e00\u79cd\u5546\u54c1\uff08\u57fa\u7840\u5927\u5b97\u54c1\uff09\u3002\u667a\u80fd\u4f53\u7684\u6709\u6548\u6027\u4e3b\u8981\u7531\u5176\u5728\u6280\u672f\u6808\u4e0a\u65b9\u7684\u6240\u6709\u4e8b\u7269\u51b3\u5b9a\uff0c\u800c\u4e0d\u662f\u7531\u667a\u80fd\u4f53\u672c\u8eab\u51b3\u5b9a\u3002<\/p>\n<p>\u8fd9\u662f\u4e00\u4e2a\u5177\u6709\u98a0\u8986\u6027\u7684\u4e3b\u5f20\uff0c\u4f46\u6709\u8bc1\u636e\u652f\u6301\u3002SWE-agent \u8bba\u6587\u901a\u8fc7\u7ecf\u9a8c\u9a8c\u8bc1\u4e86\u8fd9\u4e00\u70b9\uff1a\u76f8\u540c\u7684\u6a21\u578b\uff0c\u4ec5\u4ec5\u901a\u8fc7\u754c\u9762\u8bbe\u8ba1\u5c31\u83b7\u5f97\u4e86 64% \u7684\u6027\u80fd\u63d0\u5347\u3002OpenAI \u7684 Codex \u56e2\u961f\u5728\u5b9e\u9645\u64cd\u4f5c\u4e2d\u8bc1\u660e\u4e86\u8fd9\u4e00\u70b9\uff1a\u771f\u6b63\u91cd\u8981\u7684\u5de5\u7a0b\u5de5\u4f5c\u662f\u73af\u5883\u8bbe\u8ba1\uff0c\u800c\u4e0d\u662f\u6267\u884c\u3002Anthropic \u7684\u6846\u67b6\u5de5\u7a0b\u5de5\u4f5c\u5728\u5b9e\u8df5\u4e2d\u8bc1\u660e\u4e86\u8fd9\u4e00\u70b9\uff1a\u521d\u59cb\u5316\u667a\u80fd\u4f53\u7684\u8bbe\u7f6e\u51b3\u5b9a\u4e86\u7f16\u7801\u667a\u80fd\u4f53\u7a76\u7adf\u80fd\u5426\u53d6\u5f97\u4efb\u4f55\u8fdb\u5c55\u3002<\/p>\n<h3>\u91cd\u590d\u51fa\u73b0\u7684\u8bbe\u8ba1\u6a21\u5f0f<\/h3>\n<p>\u5728\u6240\u6709\u8fd9\u4e9b\u7cfb\u7edf\u548c\u7ec4\u7ec7\u4e2d\uff0c\u51e0\u79cd\u8bbe\u8ba1\u6a21\u5f0f\u53cd\u590d\u51fa\u73b0\u3002\u8fd9\u5e76\u975e\u5de7\u5408\uff0c\u800c\u662f\u6bcf\u5f53\u8bd5\u56fe\u5927\u89c4\u6a21\u53ef\u9760\u5730\u90e8\u7f72\u667a\u80fd\u4f53\u65f6\uff0c\u9488\u5bf9\u51fa\u73b0\u7684\u5de5\u7a0b\u95ee\u9898\u6240\u63d0\u51fa\u7684\u89e3\u51b3\u65b9\u6848\u3002<\/p>\n<h4>\u6a21\u5f0f 1\uff1a\u6e10\u8fdb\u5f0f\u62ab\u9732<\/h4>\n<p>\u4e0d\u8981\u9884\u5148\u7ed9\u667a\u80fd\u4f53\u63d0\u4f9b\u5b83\u53ef\u80fd\u9700\u8981\u7684\u6240\u6709\u4fe1\u606f\u3002\u53ea\u7ed9\u5b83\u5b9a\u4f4d\u81ea\u8eab\u6240\u9700\u7684\u6700\u5c0f\u503c\uff0c\u4ee5\u53ca\u5728\u9700\u8981\u65f6\u5bfb\u627e\u66f4\u591a\u4fe1\u606f\u7684\u6307\u9488\u3002\u8fd9\u79cd\u6a21\u5f0f\u51fa\u73b0\u5728 SWE-agent \u7684\u9650\u5236\u6027\u641c\u7d22\uff08\u4e0d\u8fd4\u56de\u6240\u6709\u7ed3\u679c\uff0c\u5f3a\u5236\u667a\u80fd\u4f53\u7ec6\u5316\u641c\u7d22\uff09\u3001OpenAI \u7684\u00a0<code>docs\/<\/code>\u00a0\u67b6\u6784\uff08\u4e00\u4e2a\u6307\u5411\u66f4\u6df1\u5c42\u6b21\u4e8b\u5b9e\u7684\u7b80\u77ed\u5730\u56fe\uff09\u3001Anthropic \u7684\u542f\u52a8\u5e8f\u5217\uff08\u5148\u8bfb\u53d6\u8fdb\u5ea6\u6587\u4ef6\uff0c\u7136\u540e\u662f\u529f\u80fd\u5217\u8868\uff09\uff0c\u4ee5\u53ca\u5b9e\u73b0\u7ed3\u6784\u5316\u4e0a\u4e0b\u6587\u5206\u5c42\u7684\u5229\u7528\u5de5\u5177\u6846\u67b6\u4e2d\u3002<\/p>\n<p>\u8fd9\u79cd\u6a21\u5f0f\u7684\u8ba4\u77e5\u539f\u56e0\u662f\u4e0a\u4e0b\u6587\u662f\u6709\u9650\u8d44\u6e90\uff0c\u4e14\u667a\u80fd\u4f53\u7684\u6ce8\u610f\u529b\u5728\u5176\u4e0a\u5e76\u975e\u5747\u5300\u5206\u5e03\u3002\u63d0\u793a\u8bcd\u5f00\u5934\u5448\u73b0\u7684\u4fe1\u606f\u5177\u6709\u4e0d\u6210\u6bd4\u4f8b\u7684\u5f71\u54cd\u529b\u3002\u4e00\u4e2a\u7b80\u77ed\u3001\u805a\u7126\u7684\u5207\u5165\u70b9\u5e76\u6307\u5411\u5176\u4ed6\u5730\u65b9\u66f4\u4e30\u5bcc\u7684\u4e0a\u4e0b\u6587\uff0c\u6bd4\u4e00\u4e2a\u4f1a\u7a00\u91ca\u6ce8\u610f\u529b\u7684\u5168\u9762\u4fe1\u606f\u8f6c\u50a8\u66f4\u6709\u6548\u3002<\/p>\n<p>\u5b9e\u8df5\u4e0a\u7684\u539f\u56e0\u662f\u7ef4\u62a4\u3002\u4e00\u4e2a\u4f5c\u4e3a\u6df1\u5ea6\u6587\u6863\u5730\u56fe\u7684\u7b80\u77ed\u5207\u5165\u70b9\uff0c\u4f60\u53ef\u4ee5\u4fdd\u6301\u5176\u51c6\u786e\u6027\u3002\u800c\u4e00\u4e2a\u5305\u542b\u6240\u6709\u5185\u5bb9\u7684\u5e9e\u5927\u6587\u6863\u5f88\u5feb\u5c31\u4f1a\u8fc7\u65f6\u5e76\u4ea7\u751f\u8d1f\u9762\u5f71\u54cd\u3002<\/p>\n<h4>\u6a21\u5f0f 2\uff1aGit \u5de5\u4f5c\u6811\u9694\u79bb<\/h4>\n<p>\u4e00\u4e2a\u667a\u80fd\u4f53\uff0c\u4e00\u4e2a\u5de5\u4f5c\u6811\u3002\u8fd9\u79cd\u6a21\u5f0f\u51fa\u73b0\u5728\u6bcf\u4e2a\u4e25\u8083\u7684\u7f16\u6392\u7cfb\u7edf\u4e2d\u3002\u7406\u7531\u5f88\u7b80\u5355\uff1a\u5f53\u4f60\u6709\u591a\u4e2a\u667a\u80fd\u4f53\u5e76\u884c\u5de5\u4f5c\uff08\u6216\u5355\u4e2a\u667a\u80fd\u4f53\u6309\u987a\u5e8f\u8fd0\u884c\u4efb\u52a1\uff09\u65f6\uff0c\u4f60\u9700\u8981\u5de5\u4f5c\u6d41\u4e4b\u95f4\u7684\u9694\u79bb\u3002\u5982\u679c\u6ca1\u6709\u9694\u79bb\uff0c\u5e76\u884c\u7684\u667a\u80fd\u4f53\u4f1a\u4e92\u76f8\u8e29\u8e0f\u5bf9\u65b9\u7684\u66f4\u6539\u3002\u5373\u4f7f\u662f\u987a\u5e8f\u6267\u884c\u7684\u667a\u80fd\u4f53\uff0c\u4f60\u4e5f\u5e0c\u671b\u5728\u66f4\u6539\u5f71\u54cd\u4e3b\u4ee3\u7801\u5e93\u4e4b\u524d\uff0c\u80fd\u591f\u5728\u9694\u79bb\u7684\u73af\u5883\u4e2d\u9a8c\u8bc1\u5b83\u4eec\u3002<\/p>\n<p>Git \u5de5\u4f5c\u6811\u5728\u6587\u4ef6\u7cfb\u7edf\u5c42\u7ea7\u63d0\u4f9b\u4e86\u8fd9\u79cd\u9694\u79bb\u3002\u6bcf\u4e2a\u667a\u80fd\u4f53\u90fd\u6709\u81ea\u5df1\u7684\u5de5\u4f5c\u76ee\u5f55\u3001\u81ea\u5df1\u7684\u5206\u652f\u548c\u81ea\u5df1\u7684\u73af\u5883\u3002\u66f4\u6539\u5728\u9694\u79bb\u4e2d\u8fdb\u884c\uff0c\u5728\u9694\u79bb\u4e2d\u6d4b\u8bd5\uff0c\u5e76\u4e14\u53ea\u6709\u5728\u901a\u8fc7\u9a8c\u8bc1\u65f6\u624d\u8fdb\u884c\u5408\u5e76\u3002\u8fd9\u5c31\u662f\u73b0\u4ee3 CI\/CD \u7cfb\u7edf\u4e3a\u4eba\u7c7b\u5de5\u7a0b\u5e08\u5de5\u4f5c\u7684\u65b9\u5f0f\uff0c\u4e8b\u5b9e\u8bc1\u660e\uff0c\u8fd9\u5bf9\u4e8e\u667a\u80fd\u4f53\u7f16\u6392\u4e5f\u662f\u5b8c\u5168\u6b63\u786e\u7684\u6a21\u578b\u3002<\/p>\n<h4>\u6a21\u5f0f 3\uff1a\u89c4\u7ea6\u4f18\u5148\uff0c\u4ed3\u5e93\u4f5c\u4e3a\u552f\u4e00\u4e8b\u5b9e\u6765\u6e90<\/h4>\n<p>\u667a\u80fd\u4f53\u5bf9\u975e\u6b63\u5f0f\u77e5\u8bc6\u662f\u76f2\u76ee\u7684\u3002\u4efb\u4f55\u5b58\u5728\u4e8e Slack \u8ba8\u8bba\u4e32\u3001Google \u6587\u6863\u6216\u67d0\u4eba\u8111\u6d77\u4e2d\u7684\u4e1c\u897f\u5bf9\u667a\u80fd\u4f53\u6765\u8bf4\u90fd\u662f\u4e0d\u53ef\u89c1\u7684\u3002\u667a\u80fd\u4f53\u552f\u4e00\u80fd\u5904\u7406\u7684\u662f\u5176\u4e0a\u4e0b\u6587\u7a97\u53e3\u4e2d\u7684\u5185\u5bb9\uff0c\u800c\u8be5\u4e0a\u4e0b\u6587\u552f\u4e00\u53ef\u9760\u7684\u6765\u6e90\u662f\u4ee3\u7801\u4ed3\u5e93\u3002<\/p>\n<p>\u8fd9\u79cd\u6a21\u5f0f\u8868\u73b0\u4e3a Anthropic \u5229\u7528\u5de5\u5177\u4e2d\u7684\u529f\u80fd\u5217\u8868\u6587\u4ef6\u3001OpenAI \u7cfb\u7edf\u4e2d\u7684\u7ed3\u6784\u5316\u00a0<code>docs\/<\/code>\u00a0\u76ee\u5f55\u3001\u5404\u79cd\u5f00\u6e90\u6846\u67b6\u4e2d\u7684\u00a0<code>AGENTS.md<\/code>\u00a0\u6587\u4ef6\uff0c\u4ee5\u53ca awesome-agent-harness \u5206\u7c7b\u6cd5\u4e2d\u7684\u89c4\u7ea6\u5de5\u5177\u5c42\u3002\u5171\u540c\u70b9\u662f\uff0c\u5728\u6267\u884c\u5f00\u59cb\u4e4b\u524d\uff0c\u89c4\u7ea6\u3001\u9700\u6c42\u3001\u67b6\u6784\u51b3\u7b56\u548c\u7ea6\u675f\u5fc5\u987b\u7f16\u7801\u5230\u4ed3\u5e93\u4e2d\u673a\u5668\u53ef\u8bfb\u7684\u6587\u4ef6\u4e2d\u3002\u5982\u679c\u667a\u80fd\u4f53\u4e0d\u80fd\u4ece\u4ed3\u5e93\u4e2d\u8bfb\u53d6\u5b83\uff0c\u5b83\u5c31\u4e0d\u5b58\u5728\u3002<\/p>\n<p>\u8fd9\u5bf9\u5de5\u7a0b\u56e2\u961f\u5982\u4f55\u8bb0\u5f55\u5176\u5de5\u4f5c\u5177\u6709\u91cd\u8981\u610f\u4e49\u3002\u6587\u6863\u4e0d\u518d\u4ec5\u4ec5\u662f\u7ed9\u4eba\u7c7b\u8bfb\u8005\u770b\u7684\u3002\u5b83\u662f\u4eba\u7c7b\u610f\u56fe\u53d8\u5f97\u5bf9\u667a\u80fd\u4f53\u53ef\u8bfb\u7684\u673a\u5236\u3002\u6a21\u7cca\u3001\u9648\u65e7\u6216\u5b58\u50a8\u5728\u4ed3\u5e93\u4e4b\u5916\u7684\u6587\u6863\uff0c\u662f\u4f1a\u4e3b\u52a8\u635f\u5bb3\u667a\u80fd\u4f53\u6027\u80fd\u7684\u6587\u6863\u3002<\/p>\n<h4>\u6a21\u5f0f 4\uff1a\u673a\u68b0\u5316\u67b6\u6784\u5f3a\u5236\u6267\u884c<\/h4>\n<p>\u4eba\u5de5\u4ee3\u7801\u5ba1\u67e5\u65e0\u6cd5\u6269\u5c55\u5230\u667a\u80fd\u4f53\u9a71\u52a8\u7684\u5f00\u53d1\u3002\u5f53\u4e00\u4e2a\u667a\u80fd\u4f53\u6bcf\u5929\u80fd\u4e3a\u6bcf\u4e2a\u5de5\u7a0b\u5e08\u5f00\u542f 3.5 \u4e2a\u62c9\u53d6\u8bf7\u6c42\u65f6\uff0c\u5ba1\u67e5\u4e0d\u80fd\u4f5c\u4e3a\u7ef4\u62a4\u4ee3\u7801\u8d28\u91cf\u548c\u67b6\u6784\u5b8c\u6574\u6027\u7684\u4e3b\u8981\u673a\u5236\u3002\u89e3\u51b3\u65b9\u6848\u662f\u5c06\u67b6\u6784\u7ea6\u675f\u7f16\u7801\u4e3a\u81ea\u52a8\u8fd0\u884c\u7684\u673a\u68b0\u5316\u68c0\u67e5\u3002<\/p>\n<p>\u81ea\u5b9a\u4e49\u9759\u6001\u4ee3\u7801\u5206\u6790\u5de5\u5177\uff08Linters\uff09\u3001\u7ed3\u6784\u5316\u6d4b\u8bd5\u548c CI \u6d41\u6c34\u7ebf\u53d6\u4ee3\u4e86\u4eba\u7c7b\u9a71\u52a8\u5f00\u53d1\u4e2d\u4ee3\u7801\u5ba1\u67e5\u7684\u5927\u90e8\u5206\u5de5\u4f5c\u3002\u5176\u4f18\u52bf\u5728\u4e8e\u673a\u68b0\u5316\u68c0\u67e5\u662f\u4e00\u81f4\u7684\u3001\u5feb\u901f\u7684\uff0c\u5e76\u5728\u8fdd\u89c4\u70b9\u7acb\u5373\u63d0\u4f9b\u53cd\u9988\u3002\u4e00\u4e2a\u6355\u83b7\u5230\u67b6\u6784\u8fdd\u89c4\u5e76\u5728\u9519\u8bef\u6d88\u606f\u4e2d\u8fd4\u56de\u4fee\u590d\u6307\u4ee4\u7684 Linter\uff0c\u6bd4\u4e09\u5929\u540e\u5728\u62c9\u53d6\u8bf7\u6c42\u8bc4\u8bba\u4e2d\u53d1\u73b0\u76f8\u540c\u8fdd\u89c4\u7684\u4ee3\u7801\u5ba1\u67e5\u5458\u66f4\u6709\u6548\u3002<\/p>\n<p>\u5173\u952e\u7684\u8bbe\u8ba1\u539f\u5219\u662f\u5f3a\u5236\u6267\u884c\u201c\u4e0d\u53d8\u6027\u201d\uff0c\u800c\u975e\u5177\u4f53\u5b9e\u73b0\u3002\u4f60\u6df1\u5ea6\u5173\u6ce8\u4f9d\u8d56\u65b9\u5411\u3001\u8fb9\u754c\u8de8\u8d8a\u3001\u63a5\u53e3\u5904\u7684\u6570\u636e\u9a8c\u8bc1\u4ee5\u53ca\u547d\u540d\u548c\u7ed3\u6784\u7684\u4e00\u81f4\u6027\u3002\u4f60\u5e76\u4e0d\u5173\u5fc3\u667a\u80fd\u4f53\u4f7f\u7528\u54ea\u4e2a\u7279\u5b9a\u7684\u5e93\u6216\u51fd\u6570\u7a76\u7adf\u5982\u4f55\u62c6\u89e3\uff0c\u53ea\u8981\u5b83\u6ee1\u8db3\u884c\u4e3a\u5951\u7ea6\u5373\u53ef\u3002\u8fd9\u8d4b\u4e88\u4e86\u667a\u80fd\u4f53\u5728\u660e\u786e\u5b9a\u4e49\u7684\u7ed3\u6784\u5185\u663e\u8457\u7684\u81ea\u4e3b\u6743\u3002<\/p>\n<h4>\u6a21\u5f0f 5\uff1a\u96c6\u6210\u53cd\u9988\u5faa\u73af<\/h4>\n<p>\u6bcf\u4e00\u4e2a\u9ad8\u6027\u80fd\u7684Harness\u67b6\u6784\u90fd\u5c3d\u53ef\u80fd\u7d27\u5bc6\u5730\u95ed\u5408\u53cd\u9988\u5faa\u73af\u3002\u8bed\u6cd5\u9519\u8bef\u7531 Linters \u5728\u7f16\u8f91\u65f6\u6355\u83b7\u3002\u8fd0\u884c\u65f6\u9519\u8bef\u901a\u8fc7\u667a\u80fd\u4f53\u53ef\u4ee5\u67e5\u8be2\u7684\u53ef\u89c2\u6d4b\u6027\u5de5\u5177\u6d6e\u73b0\u3002UI \u7f3a\u9677\u901a\u8fc7\u667a\u80fd\u4f53\u53ef\u4ee5\u9a71\u52a8\u7684\u6d4f\u89c8\u5668\u81ea\u52a8\u5316\u6355\u83b7\u3002\u6d4b\u8bd5\u5931\u8d25\u8fd4\u56de\u65f6\u5e26\u6709\u5173\u4e8e\u54ea\u91cc\u51fa\u4e86\u95ee\u9898\u4ee5\u53ca\u51fa\u5728\u4f55\u5904\u7684\u4e0a\u4e0b\u6587\u3002<\/p>\n<p>\u53e6\u4e00\u79cd\u9009\u62e9\u2014\u2014\u667a\u80fd\u4f53\u7f16\u5199\u4ee3\u7801\uff0c\u4ee3\u7801\u5728\u5916\u90e8\u8fdb\u884c\u6d4b\u8bd5\u5e76\u4ea7\u751f\u5931\u8d25\u6d88\u606f\uff0c\u5e76\u5728\u968f\u540e\u7684\u4f1a\u8bdd\u4e2d\u53cd\u9988\u2014\u2014\u5219\u66f4\u6162\u3001Token \u6210\u672c\u66f4\u9ad8\uff0c\u5e76\u4e14\u66f4\u6709\u53ef\u80fd\u4ea7\u751f\u7ea7\u8054\u6545\u969c\u3002\u53cd\u9988\u5faa\u73af\u4e2d\u4efb\u4f55\u53ef\u4ee5\u7f29\u5c0f\u884c\u52a8\u4e0e\u540e\u679c\u4e4b\u95f4\u5dee\u8ddd\u7684\u70b9\uff0c\u90fd\u662f\u53ef\u4ee5\u63d0\u9ad8\u667a\u80fd\u4f53\u6027\u80fd\u7684\u70b9\u3002<\/p>\n<p>\u8fd9\u662f\u7ecf\u5178\u8f6f\u4ef6\u5de5\u7a0b\u4e2d\u201c\u5c3d\u65e9\u53d1\u73b0\u9519\u8bef\u201d\u539f\u5219\u5728\u5229\u7528\u5de5\u5177\u4e0a\u7684\u4f53\u73b0\u3002\u53d1\u73b0\u9519\u8bef\u8d8a\u65e9\uff0c\u4fee\u590d\u6210\u672c\u8d8a\u4f4e\u3002\u5bf9\u4e8e\u667a\u80fd\u4f53\u6765\u8bf4\uff0c\u8fd9\u4e00\u70b9\u66f4\u4e3a\u6709\u529b\uff0c\u56e0\u4e3a\u672a\u80fd\u7acb\u5373\u6355\u83b7\u7684\u9519\u8bef\u4f1a\u5728\u4e0a\u4e0b\u6587\u4e2d\u7d2f\u79ef\uff0c\u5e76\u964d\u4f4e\u540e\u7eed\u63a8\u7406\u7684\u8d28\u91cf\u3002<\/p>\n<h3>\u8fd9\u5bf9\u5de5\u7a0b\u5e08\u800c\u8a00\u7a76\u7adf\u610f\u5473\u7740\u4ec0\u4e48<\/h3>\n<h4>\u53ef\u8fc1\u79fb\u7684\u6280\u80fd<\/h4>\n<p>Harness\u5de5\u7a0b\u539f\u5219\u7684\u6838\u5fc3\uff0c\u662f\u5c06\u7cfb\u7edf\u601d\u7ef4\u5e94\u7528\u4e8e\u667a\u80fd\u4f53\u73af\u5883\u3002\u5b83\u8981\u6c42\u4f60\u8db3\u591f\u6df1\u5165\u5730\u7406\u89e3\u8bed\u8a00\u6a21\u578b\u7684\u8ba4\u77e5\u67b6\u6784\uff0c\u4ece\u800c\u8bbe\u8ba1\u51fa\u4e0e\u5176\u534f\u4f5c\u800c\u975e\u5bf9\u6297\u7684\u73af\u5883\u3002\u5b83\u8981\u6c42\u4f60\u4ee5\u5206\u5e03\u5f0f\u7cfb\u7edf\u5de5\u7a0b\u4e2d\u719f\u6089\u7684\u65b9\u5f0f\u53bb\u601d\u8003\u72b6\u6001\u7ba1\u7406\u3001\u53cd\u9988\u5faa\u73af\u3001\u9519\u8bef\u6062\u590d\u548c\u4e0a\u4e0b\u6587\u4f18\u5316\uff0c\u5e76\u5c06\u5176\u5e94\u7528\u4e8e\u4e00\u4e2a\u5168\u65b0\u7684\u9886\u57df\u3002<\/p>\n<p>\u5728\u8fd9\u4e2a\u65b0\u5174\u8303\u5f0f\u4e2d\uff0c\u6700\u6709\u6548\u7684\u5de5\u7a0b\u5e08\u5e76\u4e0d\u662f\u90a3\u4e9b\u62e5\u6709\u6700\u5f3a\u63d0\u793a\u8bcd\u6280\u5de7\u7684\u4eba\uff08\u5c3d\u7ba1\u63d0\u793a\u8bcd\u4e5f\u5f88\u91cd\u8981\uff09\u3002\u800c\u662f\u90a3\u4e9b\u7406\u89e3\u6574\u4e2a\u7cfb\u7edf\u5982\u4f55\u8fd0\u4f5c\u7684\u4eba\uff1a\u4e0a\u4e0b\u6587\u5982\u4f55\u6d41\u52a8\u3001\u5728\u54ea\u91cc\u4f1a\u88ab\u6c61\u67d3\u3001\u53cd\u9988\u5faa\u73af\u5982\u4f55\u6536\u7d27\u3001\u72b6\u6001\u5982\u4f55\u8de8\u4f1a\u8bdd\u4fdd\u5b58\uff0c\u4ee5\u53ca\u5982\u4f55\u5728\u4e0d\u5fae\u89c2\u7ba1\u7406\u667a\u80fd\u4f53\u884c\u4e3a\u7684\u60c5\u51b5\u4e0b\u5f3a\u5236\u6267\u884c\u7ea6\u675f\u3002<\/p>\n<p>\u4ece\u62bd\u8c61\u89d2\u5ea6\u770b\uff0c\u8fd9\u4e9b\u5e76\u4e0d\u662f\u65b0\u6280\u80fd\u3002\u5b83\u4eec\u662f\u4f18\u79c0\u8f6f\u4ef6\u5de5\u7a0b\u5e08\u5df2\u6709\u6280\u80fd\u7684\u5ef6\u4f38\uff1a\u7cfb\u7edf\u8bbe\u8ba1\u3001API \u8bbe\u8ba1\u3001\u9519\u8bef\u5904\u7406\u3001\u6d4b\u8bd5\u7b56\u7565\u3002\u65b0\u7684\u662f\u9886\u57df\uff1a\u4e3a\u8bed\u8a00\u6a21\u578b\u667a\u80fd\u4f53\u8bbe\u8ba1\u73af\u5883\uff0c\u800c\u4e0d\u662f\u4e3a\u4eba\u7c7b\u8bbe\u8ba1\u63a5\u53e3\u3002<\/p>\n<h4>\u4f60\u5e94\u8be5\u95ee\u7684\u95ee\u9898<\/h4>\n<p>\u5f53\u4f60\u6784\u5efa\u4e00\u4e2a\u667a\u80fd\u4f53\u7cfb\u7edf\u4e14\u67d0\u4e9b\u73af\u8282\u5931\u6548\u65f6\uff0c\u8fd0\u7528Harness\u5de5\u7a0b\u601d\u7ef4\u4f1a\u4ea7\u751f\u4e0e\u539f\u751f\u601d\u7ef4\u4e0d\u540c\u7684\u4e00\u7ec4\u95ee\u9898\u3002<\/p>\n<p>\u4f60\u4e0d\u518d\u95ee\u201c\u6211\u8be5\u5982\u4f55\u5199\u51fa\u66f4\u597d\u7684\u63d0\u793a\u8bcd\uff1f\u201d\uff0c\u800c\u662f\u95ee\u201c\u667a\u80fd\u4f53\u9700\u8981\u4ec0\u4e48\u5b83\u76ee\u524d\u65e0\u6cd5\u8bbf\u95ee\u7684\u4fe1\u606f\uff1f\u201d\u3002\u4f60\u4e0d\u518d\u95ee\u201c\u4e3a\u4ec0\u4e48\u6a21\u578b\u4f1a\u72af\u8fd9\u4e2a\u9519\u8bef\uff1f\u201d\uff0c\u800c\u662f\u95ee\u201c\u7f3a\u5c11\u4e86\u4ec0\u4e48\u6837\u7684\u53cd\u9988\u5faa\u73af\uff0c\u624d\u80fd\u5728\u8fd9\u4e2a\u9519\u8bef\u4f20\u64ad\u4e4b\u524d\u6355\u83b7\u5b83\uff1f\u201d\u3002\u4f60\u4e0d\u518d\u95ee\u201c\u4e3a\u4ec0\u4e48\u667a\u80fd\u4f53\u4e0d\u6309\u6211\u8bf4\u7684\u505a\uff1f\u201d\uff0c\u800c\u662f\u95ee\u201c\u73af\u5883\u4e2d\u7684\u4ec0\u4e48\u7ea6\u675f\u963b\u788d\u4e86\u667a\u80fd\u4f53\u6267\u884c\u6211\u7684\u6307\u4ee4\uff1f\u201d\u3002<\/p>\n<p>\u8fd9\u79cd\u8f6c\u53d8\u4e0d\u4ec5\u4ec5\u662f\u8bed\u4e49\u4e0a\u7684\u3002\u5b83\u6539\u53d8\u4e86\u4f60\u6295\u5165\u5de5\u7a0b\u7cbe\u529b\u7684\u65b9\u5411\u3002\u6295\u8d44\u4e8e\u4e00\u4e2a\u89e3\u51b3\u7279\u5b9a\u5931\u8d25\u6a21\u5f0f\u7684\u66f4\u597d\u63d0\u793a\u8bcd\u662f\u5c40\u90e8\u7684\u3001\u6682\u65f6\u7684\u3002\u800c\u6295\u8d44\u4e8e\u4e00\u4e2a\u80fd\u9632\u6b62\u4e00\u7c7b\u5931\u8d25\u6a21\u5f0f\u7684\u66f4\u597d\u5de5\u5177\u5219\u662f\u901a\u7528\u7684\u3001\u6c38\u4e45\u7684\u3002Harness\u6b63\u662f\u8fd9\u79cd\u6c38\u4e45\u6027\u6295\u8d44\u7684\u6240\u5728\u5730\u3002<\/p>\n<h4>\u6267\u884c\u7684\u5546\u54c1\u5316<\/h4>\n<p>\u5728 Awesome Agent Harness \u4ee3\u7801\u4ed3\u5e93\u7684\u6838\u5fc3\u8bba\u70b9\u4e2d\uff0c\u9690\u85cf\u7740\u4e00\u4e2a\u503c\u5f97\u5766\u767d\u9648\u8ff0\u7684\u4ee4\u4eba\u4e0d\u5b89\u7684\u542b\u4e49\uff1a\u5982\u679c\u6267\u884c\u5c42\u662f\u4e00\u79cd\u5546\u54c1\uff0c\u90a3\u4e48 AI \u9a71\u52a8\u5f00\u53d1\u4e2d\u7684\u957f\u671f\u7ade\u4e89\u62a4\u57ce\u6cb3\u5c31\u4e0d\u5728\u6a21\u578b\u4e2d\uff0c\u800c\u662f\u5728Harness\u4e2d\u3002<\/p>\n<p>\u8fd9\u610f\u5473\u7740\uff0c\u90a3\u4e9b\u6295\u8d44\u4e8eHarness\u5de5\u7a0b\u2014\u2014\u6784\u5efa\u811a\u624b\u67b6\u3001\u53cd\u9988\u5faa\u73af\u3001\u53ef\u89c2\u6d4b\u6027\u3001\u89c4\u7ea6\u5de5\u5177\u4ee5\u53ca\u5141\u8bb8\u667a\u80fd\u4f53\u5927\u89c4\u6a21\u53ef\u9760\u5de5\u4f5c\u7684\u7f16\u6392\u7cfb\u7edf\u2014\u2014\u7684\u7ec4\u7ec7\u548c\u4e2a\u4eba\uff0c\u5c06\u6bd4\u90a3\u4e9b\u4e3b\u8981\u5173\u6ce8\u4f7f\u7528\u54ea\u4e2a\u6a21\u578b\u6216\u5982\u4f55\u5199\u63d0\u793a\u8bcd\u7684\u4eba\u62e5\u6709\u66f4\u6301\u4e45\u7684\u4f18\u52bf\u3002<\/p>\n<p>OpenAI \u7684 Codex \u56e2\u961f\u4e3a\u4ed6\u4eec\u7279\u5b9a\u7684\u4ee3\u7801\u5e93\u548c\u9886\u57df\u6784\u5efa\u4e86\u76f8\u5f53\u4e8e\u5b9a\u5236\u5f00\u53d1\u5e73\u53f0\u7684\u7cfb\u7edf\u3002Anthropic \u6784\u5efa\u4e86\u4e00\u4e2a\u80fd\u591f\u8ba9\u590d\u6742\u5e94\u7528\u5b9e\u73b0\u6570\u6708\u589e\u91cf\u8fdb\u5c55\u7684Harness\u67b6\u6784\u3002SWE-agent \u56e2\u961f\u6784\u5efa\u4e86\u4e00\u4e2a\u8ba9\u76f8\u540c\u6a21\u578b\u4ea7\u751f 64% \u6027\u80fd\u63d0\u5347\u7684\u754c\u9762\u3002\u8fd9\u4e9b\u4f18\u52bf\u6ca1\u6709\u4e00\u4e2a\u6765\u81ea\u6a21\u578b\uff0c\u5b83\u4eec\u5168\u90e8\u6765\u81ea\u73af\u5883\u3002<\/p>\n<p>\u6a21\u578b\u662f\u201c\u601d\u8003\u8005\u201d\u3002Harness\u662f\u201c\u601d\u8003\u7684\u5bf9\u8c61\u201d\u3002\u641e\u6e05\u695a\u8fd9\u4e2a\u533a\u522b\u5c31\u662f\u8fd9\u573a\u6e38\u620f\u7684\u5168\u90e8\u3002<\/p>\n<h3>\u6784\u5efa\u4f60\u81ea\u5df1\u7684Harness<\/h3>\n<h4>\u6700\u5c0f\u53ef\u884cHarness<\/h4>\n<p>\u4f60\u4e0d\u9700\u8981\u6784\u5efa OpenAI \u90a3\u6837\u7684\u53ef\u89c2\u6d4b\u6027\u6280\u672f\u6808\u6216 Anthropic \u5b8c\u6574\u7684\u53cc\u667a\u80fd\u4f53\u67b6\u6784\uff0c\u4e5f\u80fd\u4eceHarness\u601d\u7ef4\u4e2d\u83b7\u76ca\u3002\u5bf9\u4e8e\u4e00\u4e2a\u771f\u5b9e\u9879\u76ee\u4e2d\u7684\u7f16\u7801\u667a\u80fd\u4f53\uff0c\u6700\u5c0f\u6709\u6548\u7684Harness\u5305\u542b\u5c11\u91cf\u6838\u5fc3\u7ec4\u4ef6\u3002<\/p>\n<p><strong>\u9996\u5148\uff0c\u5efa\u7acb\u4e00\u4e2a\u6301\u4e45\u7684\u8fdb\u5ea6\u6587\u4ef6\u3002<\/strong>\u667a\u80fd\u4f53\u5728\u6bcf\u6b21\u4f1a\u8bdd\u5f00\u59cb\u65f6\u8bfb\u53d6\u5b83\u4ee5\u4e86\u89e3\u4e0a\u6b21\u505a\u4e86\u4ec0\u4e48\uff0c\u5728\u4f1a\u8bdd\u7ed3\u675f\u65f6\u5199\u5165\u5b83\u4ee5\u8bb0\u5f55\u6240\u505a\u7684\u5de5\u4f5c\u3002\u8fd9\u4e00\u4e2a\u6539\u53d8\u5c31\u80fd\u9632\u6b62\u201c\u8fc7\u65e9\u5ba3\u5e03\u80dc\u5229\u201d\u7684\u5931\u8d25\u6a21\u5f0f\uff0c\u5e76\u786e\u4fdd\u8de8\u4e0a\u4e0b\u6587\u7a97\u53e3\u8fb9\u754c\u7684\u8fde\u7eed\u6027\u3002<\/p>\n<p><strong>\u589e\u52a0\u4e00\u4e2a\u7ed3\u6784\u5316\u7684\u4efb\u52a1\u5217\u8868\u3002<\/strong>\u4e0d\u662f\u5bf9\u9879\u76ee\u7684\u6a21\u7cca\u63cf\u8ff0\uff0c\u800c\u662f\u4e00\u4e2a\u5177\u4f53\u7684\u3001\u5217\u4e3e\u51fa\u7684\u3001\u53ef\u9a8c\u8bc1\u7684\u5b8c\u6210\u6807\u51c6\u6e05\u5355\u3002\u6bcf\u9879\u90fd\u5e94\u63cf\u8ff0\u4e00\u4e2a\u53ef\u4ee5\u8fdb\u884c\u7aef\u5230\u7aef\u6d4b\u8bd5\u7684\u7528\u6237\u53ef\u89c1\u884c\u4e3a\u3002\u4e3a\u6bcf\u9879\u6807\u8bb0\u72b6\u6001\uff0c\u4e14\u667a\u80fd\u4f53\u53ea\u6709\u5728\u9a8c\u8bc1\u540e\u624d\u80fd\u66f4\u65b0\u72b6\u6001\u3002\u8fd9\u80fd\u9632\u6b62\u201c\u90e8\u5206\u5b8c\u6210\u770b\u8d77\u6765\u50cf\u5df2\u5b8c\u6210\u201d\u7684\u5931\u8d25\u6a21\u5f0f\u3002<\/p>\n<p><strong>\u5c06\u5e26\u6709\u63cf\u8ff0\u6027\u63d0\u4ea4\u6d88\u606f\u7684\u7248\u672c\u63a7\u5236\u4f5c\u4e3a\u6bcf\u6b21\u4f1a\u8bdd\u7684\u4e00\u7ea7\u90e8\u5206\u3002<\/strong>\u6bcf\u6b21\u4f1a\u8bdd\u90fd\u4ee5\u4e00\u6b21\u63d0\u4ea4\u7ed3\u675f\u3002\u5728\u4ee3\u7801\u63d0\u4ea4\u4e14\u8fdb\u5ea6\u6587\u4ef6\u66f4\u65b0\u4e4b\u524d\uff0c\u667a\u80fd\u4f53\u4e0d\u5e94\u8ba4\u4e3a\u5176\u5de5\u4f5c\u5df2\u5b8c\u6210\u3002\u8fd9\u521b\u9020\u4e86\u6e05\u6670\u7684\u4ea4\u63a5\uff0c\u4f7f\u591a\u4f1a\u8bdd\u5de5\u4f5c\u53d8\u5f97\u8fde\u8d2f\u3002<\/p>\n<p><strong>\u5982\u679c\u4f60\u5728\u6784\u5efa Web \u5e94\u7528\uff0c\u589e\u52a0\u6d4f\u89c8\u5668\u81ea\u52a8\u5316\u3002<\/strong>\u00a0\u4e00\u4e2a\u53ea\u80fd\u8bfb\u4ee3\u7801\u7684\u667a\u80fd\u4f53\u4e0e\u4e00\u4e2a\u80fd\u771f\u6b63\u4f7f\u7528\u5b83\u6240\u6784\u5efa\u7684\u5e94\u7528\u7684\u667a\u80fd\u4f53\u4e4b\u95f4\u7684\u533a\u522b\uff0c\u5c31\u50cf\u4e00\u4e2a\u53ea\u80fd\u8bfb\u4ee3\u7801\u7684\u5f00\u53d1\u8005\u4e0e\u4e00\u4e2a\u80fd\u8fd0\u884c\u5e94\u7528\u7684\u5f00\u53d1\u8005\u4e4b\u95f4\u7684\u533a\u522b\u3002\u5927\u591a\u6570\u91cd\u8981\u7684 Bug \u53ea\u6709\u5728\u8fd0\u884c\u65f6\u624d\u53ef\u89c1\u3002<\/p>\n<h4>\u73af\u5883\u5ba1\u8ba1<\/h4>\n<p>\u5982\u679c\u4f60\u5df2\u7ecf\u62e5\u6709\u4e00\u4e2a\u667a\u80fd\u4f53\u7cfb\u7edf\u4e14\u5176\u8868\u73b0\u4e0d\u4f73\uff0cHarness\u5de5\u7a0b\u65b9\u6cd5\u5efa\u8bae\u8fdb\u884c\u7279\u5b9a\u7684\u8bca\u65ad\u6d41\u7a0b\u3002\u4e0e\u5176\u5bfb\u6c42\u66f4\u597d\u7684\u6a21\u578b\u6216\u66f4\u957f\u7684\u63d0\u793a\u8bcd\uff0c\u4e0d\u5982\u8fdb\u884c\u73af\u5883\u5ba1\u8ba1\u3002<\/p>\n<p>\u95ee\uff1a\u667a\u80fd\u4f53\u9700\u8981\u54ea\u4e9b\u76ee\u524d\u65e0\u6cd5\u8bbf\u95ee\u7684\u4fe1\u606f\uff1f\u5728\u4efb\u52a1\u6d41\u7684\u54ea\u4e9b\u73af\u8282\uff0c\u667a\u80fd\u4f53\u7ecf\u5e38\u5361\u4f4f\u6216\u72af\u9519\uff1f\u7f3a\u5c11\u4e86\u54ea\u4e9b\u53cd\u9988\u80fd\u8ba9\u667a\u80fd\u4f53\u81ea\u884c\u6355\u83b7\u8fd9\u4e9b\u9519\u8bef\uff1f\u4e0a\u4e0b\u6587\u5728\u54ea\u91cc\u88ab\u65e0\u5173\u4fe1\u606f\u6c61\u67d3\u4e86\uff1f\u6709\u54ea\u4e9b\u7ea6\u675f\u9700\u8981\u5f3a\u5236\u6267\u884c\uff0c\u800c\u76ee\u524d\u5374\u4f9d\u8d56\u4e8e\u667a\u80fd\u4f53\u7684\u5224\u65ad\uff1f<\/p>\n<p>\u6bcf\u4e00\u4e2a\u95ee\u9898\u90fd\u6307\u5411\u4e00\u4e2a\u5177\u4f53\u7684Harness\u6539\u8fdb\u3002\u7f3a\u5931\u7684\u4fe1\u606f\u53d8\u6210\u4e00\u4e2a\u65b0\u5de5\u5177\u6216\u4ed3\u5e93\u4e2d\u7684\u65b0\u6587\u6863\u3002\u7f3a\u5931\u7684\u53cd\u9988\u53d8\u6210\u4e00\u4e2a\u65b0\u7684\u6d4b\u8bd5\u3001Linter \u6216\u53ef\u89c2\u6d4b\u6027\u96c6\u6210\u3002\u4e0a\u4e0b\u6587\u6c61\u67d3\u53d8\u6210\u4e00\u4e2a\u65b0\u7684\u4e0a\u4e0b\u6587\u7ba1\u7406\u7b56\u7565\u3002\u672a\u5f3a\u5236\u6267\u884c\u7684\u7ea6\u675f\u53d8\u6210\u65b0\u7684\u673a\u68b0\u5316\u68c0\u67e5\u3002<\/p>\n<p>\u8fd9\u5c31\u662f\u5229\u7528\u5de5\u5177\u5f00\u53d1\u7684\u826f\u6027\u5faa\u73af\uff1a\u6bcf\u4e00\u6b21\u5931\u8d25\u90fd\u662f\u5173\u4e8e\u73af\u5883\u9700\u8981\u4ec0\u4e48\u7684\u4fe1\u53f7\uff0c\u800c\u5bf9\u73af\u5883\u7684\u6bcf\u4e00\u6b21\u6539\u8fdb\u90fd\u4f1a\u964d\u4f4e\u672a\u6765\u6240\u6709\u667a\u80fd\u4f53\u4f1a\u8bdd\u4e2d\u8be5\u5931\u8d25\u53d1\u751f\u7684\u9891\u7387\u3002<\/p>\n<h3>\u6700\u540e\u7684\u4e00\u70b9\u601d\u8003<\/h3>\n<p>\u53d8\u9769\u6027\u6280\u672f\u5728\u65e9\u671f\u9636\u6bb5\u5982\u4f55\u88ab\u8bef\u89e3\u662f\u6709\u6a21\u5f0f\u53ef\u5faa\u7684\u3002\u90a3\u4e9b\u5438\u5f15\u516c\u4f17\u6ce8\u610f\u529b\u7684\u4e1c\u897f\u2014\u2014\u539f\u751f\u80fd\u529b\u3001\u4ee4\u4eba\u5370\u8c61\u6df1\u523b\u7684\u6f14\u793a\u3001\u57fa\u51c6\u6d4b\u8bd5\u5206\u6570\u2014\u2014\u5f88\u5c11\u662f\u51b3\u5b9a\u957f\u671f\u80dc\u8d1f\u7684\u5173\u952e\u3002\u57fa\u7840\u8bbe\u65bd\u5c42\u3001Harness\u3001\u73af\u5883\uff0c\u901a\u5e38\u624d\u662f\u771f\u6b63\u4ef7\u503c\u4ea7\u751f\u548c\u88ab\u6355\u83b7\u7684\u5730\u65b9\u3002<\/p>\n<p>\u4e92\u8054\u7f51\u5177\u6709\u53d8\u9769\u6027\uff0c\u4e0d\u662f\u56e0\u4e3a HTML \u7684\u5b58\u5728\uff0c\u800c\u662f\u56e0\u4e3a\u641c\u7d22\u5f15\u64ce\u548c\u6d4f\u89c8\u5668\u8ba9\u4e92\u8054\u7f51\u53d8\u5f97\u53ef\u5bfc\u822a\u3002\u79fb\u52a8\u7aef\u5177\u6709\u53d8\u9769\u6027\uff0c\u4e0d\u662f\u56e0\u4e3a\u667a\u80fd\u624b\u673a\u7684\u5b58\u5728\uff0c\u800c\u662f\u56e0\u4e3a\u5e94\u7528\u5546\u5e97\u548c\u5f00\u53d1\u5de5\u5177\u4f7f\u5f97\u5728\u667a\u80fd\u624b\u673a\u4e0a\u8fdb\u884c\u5927\u89c4\u6a21\u5f00\u53d1\u5e94\u7528\u6210\u4e3a\u53ef\u80fd\u3002\u5728\u8fd9\u4e24\u4e2a\u6848\u4f8b\u4e2d\uff0c\u7ec4\u7ec7\u5e95\u5c42\u80fd\u529b\u7684\u5e73\u53f0\u5c42\u624d\u662f\u6301\u4e45\u4ef7\u503c\u7684\u6240\u5728\u3002<\/p>\n<p>AI \u667a\u80fd\u4f53\u6b63\u9075\u5faa\u540c\u6837\u7684\u6a21\u5f0f\u3002\u80fd\u529b\u5df2\u7ecf\u5b58\u5728\u3002\u95ee\u9898\u5728\u4e8e\u8c01\u80fd\u6784\u5efa\u51fa\u8ba9\u8fd9\u79cd\u80fd\u529b\u53d8\u5f97\u53ef\u9760\u3001\u53ef\u63a7\u4e14\u53ef\u6301\u7eed\u6539\u8fdb\u7684\u73af\u5883\u3002SWE-agent \u7684\u7814\u7a76\u4eba\u5458\u5728 2024 \u5e74\u7406\u89e3\u4e86\u8fd9\u4e00\u70b9\u5e76\u8fdb\u884c\u4e86\u91cf\u5316\u6f14\u793a\u3002Anthropic \u5728\u6784\u5efa Claude Code \u65f6\u7406\u89e3\u4e86\u8fd9\u4e00\u70b9\u5e76\u516c\u5f00\u8bb0\u5f55\u3002OpenAI \u5728\u6784\u5efa\u5176\u5185\u90e8\u4ea7\u54c1\u65f6\u7406\u89e3\u4e86\u8fd9\u4e00\u70b9\u5e76\u5206\u4eab\u4e86\u6559\u8bad\u3002Awesome Agent Harness \u793e\u533a\u6b63\u5728\u901a\u8fc7\u6570\u5341\u79cd\u5de5\u5177\u548c\u6846\u67b6\u4e2d\u5bf9\u5176\u8fdb\u884c\u5f52\u7c7b\u3002<\/p>\n<p><strong>Harness\u5c31\u662f\u4e00\u5207\u3002<\/strong>\u6a21\u578b\u662f\u63a8\u7406\u5f15\u64ce\u3002Harness\u662f\u4e0a\u4e0b\u6587\u3001\u7ea6\u675f\u3001\u53cd\u9988\u5faa\u73af\u3001\u5185\u5b58\u3001\u5de5\u5177\u548c\u811a\u624b\u67b6\uff0c\u5b83\u4eec\u51b3\u5b9a\u4e86\u63a8\u7406\u5f15\u64ce\u7a76\u7adf\u80fd\u6210\u5c31\u4ec0\u4e48\u3002<\/p>\n<p>\u641e\u597dHarness\u4e0d\u662f\u4e00\u4e2a\u63d0\u793a\u8bcd\u5de5\u7a0b\u95ee\u9898\uff0c\u5b83\u662f\u4e00\u4e2a\u7cfb\u7edf\u5de5\u7a0b\u95ee\u9898\u3002\u800c\u4e14\u5b83\u662f\u76ee\u524d\u5e94\u7528 AI \u9886\u57df\u6700\u91cd\u8981\u7684\u5de5\u7a0b\u95ee\u9898\u3002<\/p>\n<p>\u636e\u6b64\u6784\u5efa\u5427\u3002<\/p>\n<p>\u53c2\u8003\u94fe\u63a5\uff1ahttps:\/\/x.com\/rohit4verse\/status\/2033945654377283643?s=46<\/p>\n<p>\u6587\u7ae0\u6765\u81ea\uff1a51CTO<\/p>\n<\/div>\n<\/div>\n<div class=\"pvc_clear\"><\/div>\n<p id=\"pvc_stats_29087\" class=\"pvc_stats total_only  \" data-element-id=\"29087\" style=\"\"><i class=\"pvc-stats-icon medium\" aria-hidden=\"true\"><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" version=\"1.0\" viewBox=\"0 0 502 315\" preserveAspectRatio=\"xMidYMid meet\"><g transform=\"translate(0,332) scale(0.1,-0.1)\" fill=\"\" stroke=\"none\"><path d=\"M2394 3279 l-29 -30 -3 -207 c-2 -182 0 -211 15 -242 39 -76 157 -76 196 0 15 31 17 60 15 243 l-3 209 -33 29 c-26 23 -41 29 -80 29 -41 0 -53 -5 -78 -31z\"\/><path d=\"M3085 3251 c-45 -19 -58 -50 -96 -229 -47 -217 -49 -260 -13 -295 52 -53 146 -42 177 20 16 31 87 366 87 410 0 70 -86 122 -155 94z\"\/><path d=\"M1751 3234 c-13 -9 -29 -31 -37 -50 -12 -29 -10 -49 21 -204 19 -94 39 -189 45 -210 14 -50 54 -80 110 -80 34 0 48 6 76 34 21 21 34 44 34 59 0 14 -18 113 -40 219 -37 178 -43 195 -70 221 -36 32 -101 37 -139 11z\"\/><path d=\"M1163 3073 c-36 -7 -73 -59 -73 -102 0 -56 133 -378 171 -413 34 -32 83 -37 129 -13 70 36 67 87 -16 290 -86 209 -89 214 -129 231 -35 14 -42 15 -82 7z\"\/><path d=\"M3689 3066 c-15 -9 -33 -30 -42 -48 -48 -103 -147 -355 -147 -375 0 -98 131 -148 192 -74 13 15 57 108 97 206 80 196 84 226 37 273 -30 30 -99 39 -137 18z\"\/><path d=\"M583 2784 c-38 -19 -67 -74 -58 -113 9 -42 211 -354 242 -373 16 -10 45 -18 66 -18 51 0 107 52 107 100 0 39 -1 41 -124 234 -80 126 -108 162 -133 173 -41 17 -61 16 -100 -3z\"\/><path d=\"M4250 2784 c-14 -9 -74 -91 -133 -183 -95 -150 -107 -173 -107 -213 0 -55 33 -94 87 -104 67 -13 90 8 211 198 130 202 137 225 78 284 -27 27 -42 34 -72 34 -22 0 -50 -8 -64 -16z\"\/><path d=\"M2275 2693 c-553 -48 -1095 -270 -1585 -649 -135 -104 -459 -423 -483 -476 -23 -49 -22 -139 2 -186 73 -142 361 -457 571 -626 285 -228 642 -407 990 -497 242 -63 336 -73 660 -74 310 0 370 5 595 52 535 111 1045 392 1455 803 122 121 250 273 275 326 19 41 19 137 0 174 -41 79 -309 363 -465 492 -447 370 -946 591 -1479 653 -113 14 -422 18 -536 8z m395 -428 c171 -34 330 -124 456 -258 112 -119 167 -219 211 -378 27 -96 24 -300 -5 -401 -72 -255 -236 -447 -474 -557 -132 -62 -201 -76 -368 -76 -167 0 -236 14 -368 76 -213 98 -373 271 -451 485 -162 444 86 934 547 1084 153 49 292 57 452 25z m909 -232 c222 -123 408 -262 593 -441 76 -74 138 -139 138 -144 0 -16 -233 -242 -330 -319 -155 -123 -309 -223 -461 -299 l-81 -41 32 46 c18 26 49 83 70 128 143 306 141 649 -6 957 -25 52 -61 116 -79 142 l-34 47 45 -20 c26 -10 76 -36 113 -56z m-2057 25 c-40 -58 -105 -190 -130 -263 -110 -324 -59 -707 132 -981 25 -35 42 -64 37 -64 -19 0 -241 119 -326 174 -188 122 -406 314 -532 468 l-58 71 108 103 c185 178 428 349 672 473 66 33 121 60 123 61 2 0 -10 -19 -26 -42z\"\/><path d=\"M2375 1950 c-198 -44 -350 -190 -395 -379 -18 -76 -8 -221 19 -290 114 -284 457 -406 731 -260 98 52 188 154 231 260 27 69 37 214 19 290 -38 163 -166 304 -326 360 -67 23 -215 33 -279 19z\"\/><\/g><\/svg><\/i> <img loading=\"lazy\" decoding=\"async\" width=\"16\" height=\"16\" alt=\"Loading\" src=\"https:\/\/aif.amtbbs.org\/wp-content\/plugins\/page-views-count\/ajax-loader-2x.gif\" border=0 \/><\/p>\n<div class=\"pvc_clear\"><\/div>\n","protected":false},"excerpt":{"rendered":"<p>\u53d8\u9769\u6027\u6280\u672f\u5728\u65e9\u671f\u9636\u6bb5\u5982\u4f55\u88ab\u8bef\u89e3\u662f\u6709\u6a21\u5f0f\u53ef\u5faa\u7684\u3002\u90a3\u4e9b\u5438\u5f15\u516c\u4f17\u6ce8\u610f\u529b\u7684\u4e1c\u897f\u2014\u2014\u539f\u751f\u80fd\u529b\u3001\u4ee4\u4eba\u5370\u8c61\u6df1\u523b\u7684\u6f14\u793a\u3001\u57fa\u51c6\u6d4b [&hellip;]<\/p>\n<div class=\"pvc_clear\"><\/div>\n<p id=\"pvc_stats_29087\" class=\"pvc_stats total_only  \" data-element-id=\"29087\" style=\"\"><i class=\"pvc-stats-icon medium\" aria-hidden=\"true\"><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" version=\"1.0\" viewBox=\"0 0 502 315\" preserveAspectRatio=\"xMidYMid meet\"><g transform=\"translate(0,332) scale(0.1,-0.1)\" fill=\"\" stroke=\"none\"><path d=\"M2394 3279 l-29 -30 -3 -207 c-2 -182 0 -211 15 -242 39 -76 157 -76 196 0 15 31 17 60 15 243 l-3 209 -33 29 c-26 23 -41 29 -80 29 -41 0 -53 -5 -78 -31z\"\/><path d=\"M3085 3251 c-45 -19 -58 -50 -96 -229 -47 -217 -49 -260 -13 -295 52 -53 146 -42 177 20 16 31 87 366 87 410 0 70 -86 122 -155 94z\"\/><path d=\"M1751 3234 c-13 -9 -29 -31 -37 -50 -12 -29 -10 -49 21 -204 19 -94 39 -189 45 -210 14 -50 54 -80 110 -80 34 0 48 6 76 34 21 21 34 44 34 59 0 14 -18 113 -40 219 -37 178 -43 195 -70 221 -36 32 -101 37 -139 11z\"\/><path d=\"M1163 3073 c-36 -7 -73 -59 -73 -102 0 -56 133 -378 171 -413 34 -32 83 -37 129 -13 70 36 67 87 -16 290 -86 209 -89 214 -129 231 -35 14 -42 15 -82 7z\"\/><path d=\"M3689 3066 c-15 -9 -33 -30 -42 -48 -48 -103 -147 -355 -147 -375 0 -98 131 -148 192 -74 13 15 57 108 97 206 80 196 84 226 37 273 -30 30 -99 39 -137 18z\"\/><path d=\"M583 2784 c-38 -19 -67 -74 -58 -113 9 -42 211 -354 242 -373 16 -10 45 -18 66 -18 51 0 107 52 107 100 0 39 -1 41 -124 234 -80 126 -108 162 -133 173 -41 17 -61 16 -100 -3z\"\/><path d=\"M4250 2784 c-14 -9 -74 -91 -133 -183 -95 -150 -107 -173 -107 -213 0 -55 33 -94 87 -104 67 -13 90 8 211 198 130 202 137 225 78 284 -27 27 -42 34 -72 34 -22 0 -50 -8 -64 -16z\"\/><path d=\"M2275 2693 c-553 -48 -1095 -270 -1585 -649 -135 -104 -459 -423 -483 -476 -23 -49 -22 -139 2 -186 73 -142 361 -457 571 -626 285 -228 642 -407 990 -497 242 -63 336 -73 660 -74 310 0 370 5 595 52 535 111 1045 392 1455 803 122 121 250 273 275 326 19 41 19 137 0 174 -41 79 -309 363 -465 492 -447 370 -946 591 -1479 653 -113 14 -422 18 -536 8z m395 -428 c171 -34 330 -124 456 -258 112 -119 167 -219 211 -378 27 -96 24 -300 -5 -401 -72 -255 -236 -447 -474 -557 -132 -62 -201 -76 -368 -76 -167 0 -236 14 -368 76 -213 98 -373 271 -451 485 -162 444 86 934 547 1084 153 49 292 57 452 25z m909 -232 c222 -123 408 -262 593 -441 76 -74 138 -139 138 -144 0 -16 -233 -242 -330 -319 -155 -123 -309 -223 -461 -299 l-81 -41 32 46 c18 26 49 83 70 128 143 306 141 649 -6 957 -25 52 -61 116 -79 142 l-34 47 45 -20 c26 -10 76 -36 113 -56z m-2057 25 c-40 -58 -105 -190 -130 -263 -110 -324 -59 -707 132 -981 25 -35 42 -64 37 -64 -19 0 -241 119 -326 174 -188 122 -406 314 -532 468 l-58 71 108 103 c185 178 428 349 672 473 66 33 121 60 123 61 2 0 -10 -19 -26 -42z\"\/><path d=\"M2375 1950 c-198 -44 -350 -190 -395 -379 -18 -76 -8 -221 19 -290 114 -284 457 -406 731 -260 98 52 188 154 231 260 27 69 37 214 19 290 -38 163 -166 304 -326 360 -67 23 -215 33 -279 19z\"\/><\/g><\/svg><\/i> <img loading=\"lazy\" decoding=\"async\" width=\"16\" height=\"16\" alt=\"Loading\" src=\"https:\/\/aif.amtbbs.org\/wp-content\/plugins\/page-views-count\/ajax-loader-2x.gif\" border=0 \/><\/p>\n<div class=\"pvc_clear\"><\/div>\n","protected":false},"author":56,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[3,23,20,80],"tags":[1698,2131,2130],"class_list":["post-29087","post","type-post","status-publish","format-standard","hentry","category-ai","category-23","category-20","category-80","tag-cursor","tag-openai","tag-claudecode"],"_links":{"self":[{"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/posts\/29087","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/users\/56"}],"replies":[{"embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/comments?post=29087"}],"version-history":[{"count":3,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/posts\/29087\/revisions"}],"predecessor-version":[{"id":29101,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/posts\/29087\/revisions\/29101"}],"wp:attachment":[{"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/media?parent=29087"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/categories?post=29087"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/tags?post=29087"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}