[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7EX4HPZPE52J4FB9AVG2FQC":3,"skill-same:01M7EX4HPZPE52J4FB9AVG2FQC":60},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":38,"commitSha":39,"skillDirectory":40,"forksCount":41,"repositoryHtmlUrl":42,"repositoryHomepageUrl":6,"ownerHtmlUrl":43,"ownerBlogUrl":6,"ownerDescription":44,"updatedAt":45,"install":46,"resources":59},"01M7EX4HPZPE52J4FB9AVG2FQC","auto-research",null,"Research uncertain questions with an explicit, user-approved web search or ChatGPT consultation, then present options and wait for implementation approval.","lza6\u002FArmor-piercing-system-prompts","lza6","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F128880206?v=4","01M7EB3K7WPX5T28EJ2NVHE6EG","Chinese-tingfeng","01M7EB3K97E1KQ16QWKJMFNGX8",11,"MIT","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F05-技能与代理开发\u002Fauto-research\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: auto-research\ndescription: Research uncertain questions with an explicit, user-approved web search or ChatGPT consultation, then present options and wait for implementation approval.\ncategory: automation\nrisk: critical\nsource: self\nsource_type: self\ndate_added: \"2026-07-09\"\nauthor: zyu51\ntags: [research, chatgpt, playwright, browser-automation, decision-support, chinese]\ntools: [claude, playwright]\nlicense: MIT\n---\n\n# Auto-Research Skill\n\n## Overview\n\nWhen implementing tasks, Claude Code can encounter uncertainties — design choices, algorithm details, API usage, or best practices. This skill provides an explicit-consent research path, presents findings, and waits for user approval before writing code.\n\nThe skill supports web research and an optional ChatGPT consultation. It never sends\nconversation context, files, browser state, or credentials to a third party without the\nuser's explicit approval of the exact, redacted text.\n\n## When to Use This Skill\n\n- User asks a question where multiple valid approaches exist\n- Claude is uncertain about algorithm details or API usage\n- Design\u002Farchitecture choices need comparison\n- The user explicitly asks to search the web or consult ChatGPT and approves the proposed query\n\n## How It Works\n\n**Step 1: Propose the research boundary** — State the source to use, the exact query or\nredacted prompt, whether any local\u002Fworkspace text would leave the machine, and the likely\ncost. Wait for the user to approve that exact boundary.\n\n**Step 2: Research** — After approval, use web search or a browser session the user has\nexplicitly selected and authorized. Use a pinned, user-configured browser automation\nconnector; do not install packages automatically, use `@latest`, or access browser cookies,\nother tabs, saved passwords, or sessions.\n\n**Step 3: Present** — Distill findings into concise options with sources, presented to the user.\n\n**Step 4: Await Approval** — Do NOT write code until the user says \"go ahead\" or picks an option.\n\n**Step 5: Implement** — Once approved, execute with confidence.\n\n### Explicit ChatGPT Consultation\n\nDo not treat `?`, `??`, or another shorthand as consent. First propose a minimal prompt,\nfor example: `请评估这个已脱敏的方案的正确性、完整性和可改进之处：\u003Ctext>`.\nExplicitly identify every piece of text that would be sent. Only after the user confirms\nthe exact prompt may you open the selected ChatGPT session, submit that prompt, and present\nthe response. Do not include conversation history by default.\n\nRedact secrets, personal data, proprietary code, customer data, and internal URLs before\nproposing the prompt. If safe redaction is not possible, do not submit it.\n\n### Browser Automation Boundary\n\nIf browser automation is necessary, the user must separately authorize the selected browser\nprofile and connector version. Restrict the session to the consultation tab. Do not inspect,\nreuse, export, or rely on cookies from other tabs or profiles.\n\n## Examples\n\n### Example 1: Design Question with GPT\n```\nUser: PyTorch 中自定义 ADMM 优化器怎么设计？\nClaude: 我可以搜索公开资料，或将以下已脱敏问题发给 ChatGPT：\n        “如何设计 PyTorch 自定义 ADMM 优化器？请比较可行模式。”\n        不会发送工作区文件或对话历史。是否允许？\nUser: 允许发送这段文字\nClaude: [Opens only the authorized consultation tab, submits the approved prompt]\nClaude: GPT suggests approach A with these pros\u002Fcons. Proceed?\nUser: 行\nClaude: [Implements code]\n```\n\n### Example 2: Web Search\n```\nUser: ?? ADMM convergence criteria best practices\nClaude: I can search public sources for the exact redacted query\n        “ADMM convergence criteria best practices”. No workspace files or conversation\n        history will be sent. May I send that text to WebSearch and fetch the results?\nUser: Yes, send that query\nClaude: [WebSearch + WebFetch → finds Boyd et al. paper, extracts criteria]\nClaude: Boyd recommends ||r|| \u003C ε·max(||Ax||, ||Bz||, ||c||). Use this?\nUser: Yes\nClaude: [Implements]\n```\n\n## Best Practices\n- ✅ Always present findings to user before writing code\n- ✅ Use `page.fill()` for instant text injection instead of `keyboard.type()`\n- ✅ Ask for fresh approval before every external consultation\n- ✅ Include sources in findings\n- ❌ Don't skip research and write code speculatively\n- ❌ Don't send context, files, or browser data because of a shorthand trigger\n- ❌ Don't alter the user's browser profile or session state\n\n## Limitations\n- Requires a user-configured, pinned browser automation connector if browser consultation is used\n- ChatGPT consultation is optional; use ordinary web search when it meets the need\n- GPT response time varies (10-30s typically)\n- Web search quality depends on available sources\n- Does not replace expert domain knowledge — always let user make the final call\n\n## Security & Safety Notes\n- Obtain explicit consent for each third-party submission, including the exact redacted text\n- Never access, export, or depend on cookies, saved passwords, or unrelated browser tabs\n- Never submit sensitive credentials, tokens, proprietary code, personal data, or internal URLs\n- Do not install or execute browser tooling from an unpinned package version\n\n## Common Pitfalls\n\n| Problem | Solution |\n|---------|----------|\n| ChatGPT shows login page | Let the user log in themselves; do not handle cookies or credentials |\n| The prompt contains sensitive context | Redact it or use local reasoning instead |\n| Browser automation is unavailable | Use web search or stop and ask the user for a different approved method |\n\n## Related Skills\n- @systematic-debugging — use when debugging Playwright interactions with ChatGPT\n- @condition-based-waiting — use when waiting for GPT responses in the browser\n","# Auto-Research Skill\n\n## Overview\n\nWhen implementing tasks, Claude Code can encounter uncertainties — design choices, algorithm details, API usage, or best practices. This skill provides an explicit-consent research path, presents findings, and waits for user approval before writing code.\n\nThe skill supports web research and an optional ChatGPT consultation. It never sends\nconversation context, files, browser state, or credentials to a third party without the\nuser's explicit approval of the exact, redacted text.\n\n## When to Use This Skill\n\n- User asks a question where multiple valid approaches exist\n- Claude is uncertain about algorithm details or API usage\n- Design\u002Farchitecture choices need comparison\n- The user explicitly asks to search the web or consult ChatGPT and approves the proposed query\n\n## How It Works\n\n**Step 1: Propose the research boundary** — State the source to use, the exact query or\nredacted prompt, whether any local\u002Fworkspace text would leave the machine, and the likely\ncost. Wait for the user to approve that exact boundary.\n\n**Step 2: Research** — After approval, use web search or a browser session the user has\nexplicitly selected and authorized. Use a pinned, user-configured browser automation\nconnector; do not install packages automatically, use `@latest`, or access browser cookies,\nother tabs, saved passwords, or sessions.\n\n**Step 3: Present** — Distill findings into concise options with sources, presented to the user.\n\n**Step 4: Await Approval** — Do NOT write code until the user says \"go ahead\" or picks an option.\n\n**Step 5: Implement** — Once approved, execute with confidence.\n\n### Explicit ChatGPT Consultation\n\nDo not treat `?`, `??`, or another shorthand as consent. First propose a minimal prompt,\nfor example: `请评估这个已脱敏的方案的正确性、完整性和可改进之处：\u003Ctext>`.\nExplicitly identify every piece of text that would be sent. Only after the user confirms\nthe exact prompt may you open the selected ChatGPT session, submit that prompt, and present\nthe response. Do not include conversation history by default.\n\nRedact secrets, personal data, proprietary code, customer data, and internal URLs before\nproposing the prompt. If safe redaction is not possible, do not submit it.\n\n### Browser Automation Boundary\n\nIf browser automation is necessary, the user must separately authorize the selected browser\nprofile and connector version. Restrict the session to the consultation tab. Do not inspect,\nreuse, export, or rely on cookies from other tabs or profiles.\n\n## Examples\n\n### Example 1: Design Question with GPT\n```\nUser: PyTorch 中自定义 ADMM 优化器怎么设计？\nClaude: 我可以搜索公开资料，或将以下已脱敏问题发给 ChatGPT：\n        “如何设计 PyTorch 自定义 ADMM 优化器？请比较可行模式。”\n        不会发送工作区文件或对话历史。是否允许？\nUser: 允许发送这段文字\nClaude: [Opens only the authorized consultation tab, submits the approved prompt]\nClaude: GPT suggests approach A with these pros\u002Fcons. Proceed?\nUser: 行\nClaude: [Implements code]\n```\n\n### Example 2: Web Search\n```\nUser: ?? ADMM convergence criteria best practices\nClaude: I can search public sources for the exact redacted query\n        “ADMM convergence criteria best practices”. No workspace files or conversation\n        history will be sent. May I send that text to WebSearch and fetch the results?\nUser: Yes, send that query\nClaude: [WebSearch + WebFetch → finds Boyd et al. paper, extracts criteria]\nClaude: Boyd recommends ||r|| \u003C ε·max(||Ax||, ||Bz||, ||c||). Use this?\nUser: Yes\nClaude: [Implements]\n```\n\n## Best Practices\n- ✅ Always present findings to user before writing code\n- ✅ Use `page.fill()` for instant text injection instead of `keyboard.type()`\n- ✅ Ask for fresh approval before every external consultation\n- ✅ Include sources in findings\n- ❌ Don't skip research and write code speculatively\n- ❌ Don't send context, files, or browser data because of a shorthand trigger\n- ❌ Don't alter the user's browser profile or session state\n\n## Limitations\n- Requires a user-configured, pinned browser automation connector if browser consultation is used\n- ChatGPT consultation is optional; use ordinary web search when it meets the need\n- GPT response time varies (10-30s typically)\n- Web search quality depends on available sources\n- Does not replace expert domain knowledge — always let user make the final call\n\n## Security & Safety Notes\n- Obtain explicit consent for each third-party submission, including the exact redacted text\n- Never access, export, or depend on cookies, saved passwords, or unrelated browser tabs\n- Never submit sensitive credentials, tokens, proprietary code, personal data, or internal URLs\n- Do not install or execute browser tooling from an unpinned package version\n\n## Common Pitfalls\n\n| Problem | Solution |\n|---------|----------|\n| ChatGPT shows login page | Let the user log in themselves; do not handle cookies or credentials |\n| The prompt contains sensitive context | Redact it or use local reasoning instead |\n| Browser automation is unavailable | Use web search or stop and ask the user for a different approved method |\n\n## Related Skills\n- @systematic-debugging — use when debugging Playwright interactions with ChatGPT\n- @condition-based-waiting — use when waiting for GPT responses in the browser\n",{"name":5,"risk":24,"tags":25,"tools":32,"author":34,"source":35,"license":15,"category":36,"date_added":37,"description":7,"source_type":35},"critical",[26,27,28,29,30,31],"research","chatgpt","playwright","browser-automation","decision-support","chinese",[33,28],"claude","zyu51","self","automation","2026-07-09","main","d69096e239f2a89aabc2cd078c47664a9d3b1e11","05-技能与代理开发\u002Fauto-research",3,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts","https:\u002F\u002Fgithub.com\u002Flza6","个人vx：Tf00798\r\n个人✈：@TFGY999","2026-10-08 14:29:22",{"source":8,"skillName":5,"commands":47,"prompt":57,"zipAvailable":58},[48,51,54],{"tool":49,"command":50},"npx","npx skills add lza6\u002FArmor-piercing-system-prompts --skill auto-research",{"tool":52,"command":53},"bunx","bunx skills add lza6\u002FArmor-piercing-system-prompts --skill auto-research",{"tool":55,"command":56},"pnpm","pnpm dlx skills add lza6\u002FArmor-piercing-system-prompts --skill auto-research","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F05-技能与代理开发\u002Fauto-research\u002FSKILL.md",true,[],[61,69,74,81,88,95],{"skillNo":62,"name":63,"nameZh":6,"description":64,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":65,"categories":66,"keywords":67,"publishedAt":68},"01M7FFG485KY83TD6HSC4YPCMC","obsidian-clipper-template-creator","Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobsidian-clipper-template-creator\u002FSKILL.md",[],[],"2026-11-16 10:05:34",{"skillNo":70,"name":63,"nameZh":6,"description":64,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":71,"categories":72,"keywords":73,"publishedAt":68},"01M7FFG0T6AZ0STEJPSB3CX84K","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobsidian-clipper-template-creator-antigravity-awesome-skills-main\u002FSKILL.md",[],[],{"skillNo":75,"name":76,"nameZh":6,"description":77,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"publishedAt":68},"01M7FFFY0QEM14SKWPDYGMXM2P","python-api","Programmatic Object-Detection-Metrics API guidance for BoundingBox, BoundingBoxes, Evaluator, enums, coordinate conversion, AP\u002FmAP semantics, and source-style import diagnostics.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002Fsub-skills\u002Fpython-api\u002FSKILL.md",[],[],{"skillNo":82,"name":83,"nameZh":6,"description":84,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":85,"categories":86,"keywords":87,"publishedAt":68},"01M7FFFVZQZ2GJ0C0TGP7JHD20","file-evaluation","Folder-based PASCAL VOC AP\u002FmAP evaluation from ground-truth and detection text files.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002Fsub-skills\u002Ffile-evaluation\u002FSKILL.md",[],[],{"skillNo":89,"name":90,"nameZh":6,"description":91,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":92,"categories":93,"keywords":94,"publishedAt":68},"01M7FFFSFNYSE77SJ6PE1CK2DA","object-detection-metrics","Guides legacy Object-Detection-Metrics PASCAL VOC AP\u002FmAP evaluation from detection text folders and direct Python API classes.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002FSKILL.md",[],[],{"skillNo":96,"name":97,"nameZh":6,"description":98,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":99,"categories":100,"keywords":101,"publishedAt":68},"01M7FFFP0GCK1RVXBCPQ8APBMT","nx-workspace-patterns","Configure and optimize Nx monorepo workspaces. Use when setting up Nx, configuring project boundaries, optimizing build caching, or implementing affected commands.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fnx-workspace-patterns\u002FSKILL.md",[],[]]