[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7EVT90GV8S9PPKTX2N7RFSJ":3,"skill-same:01M7EVT90GV8S9PPKTX2N7RFSJ":48},{"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":28,"commitSha":29,"skillDirectory":30,"forksCount":14,"repositoryHtmlUrl":31,"repositoryHomepageUrl":31,"ownerHtmlUrl":32,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":33,"install":34,"resources":47},"01M7EVT90GV8S9PPKTX2N7RFSJ","efficient-web-research",null,"Protocol for token-efficient web research. Use when accessing URLs, GitHub repos, or running search queries. Prevents full-page fetching waste.","Mr-Nobody-Anonymous\u002FAll-skills","Mr-Nobody-Anonymous","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F184253119?v=4","01M7AEXFJHKTYQW5R0ZCQQTWFV","anonymous","01M7C00M3GQR4QKP9GZA42GFAK",0,"MIT","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fdevelopment\u002Fefficient-web-research\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\ncategory: development\nname: efficient-web-research\nsource: community\ndate_added: \"2026-09-04\"\nrisk: safe\ndescription: >\n  Protocol for token-efficient web research. Use when accessing URLs, GitHub repos, or running search queries. Prevents full-page fetching waste.\n---\n\n# Efficient Web Research Skill\n\nA protocol for accessing web content in the most token-efficient, accurate, and structured way —\nusing the right tool at the right depth, and stopping as soon as the question is answerable.\n\n---\n\n## When to Use\n- Use this skill when the task matches this description: Protocol for token-efficient web research. Use when accessing URLs, GitHub repos, or running search queries. Prevents full-page fetching waste.\n\n## Core Principle\n\n> **Fetch the minimum needed to answer. Skim before you dive. Stop when you can answer.**\n\nEvery unnecessary fetch wastes tokens and adds noise. This skill enforces a layered approach\nwhere you escalate fetch depth only when shallower layers fail.\n\n---\n\n## Step 1 — Classify the Input\n\nBefore fetching anything, identify what kind of input you received:\n\n| Input Type | Example | Go To |\n|---|---|---|\n| GitHub repo URL | `github.com\u002Fuser\u002Frepo` | [GitHub Protocol](#github-protocol) |\n| Specific page URL | `docs.python.org\u002F3\u002Flibrary\u002Fos` | [URL Protocol](#url-protocol) |\n| Topic \u002F query (no URL) | \"how does RAFT consensus work\" | [Search Protocol](#search-protocol) |\n| Multiple URLs | List of links | [Multi-URL Protocol](#multi-url-protocol) |\n| PDF \u002F file link | `.pdf`, `.txt`, `.md` URL | [File Protocol](#file-protocol) |\n\n---\n\n## GitHub Protocol\n\nUse when input is a GitHub URL (repo, file, PR, issue, etc.)\n\n### Step 1 — Parse the URL\n\n```\ngithub.com\u002F{owner}\u002F{repo}                → Repo root\ngithub.com\u002F{owner}\u002F{repo}\u002Ftree\u002F{branch}  → Directory\ngithub.com\u002F{owner}\u002F{repo}\u002Fblob\u002F{branch}\u002F{path} → Single file\ngithub.com\u002F{owner}\u002F{repo}\u002Fissues\u002F{n}     → Issue\ngithub.com\u002F{owner}\u002F{repo}\u002Fpull\u002F{n}       → Pull request\n```\n\n### Step 2 — Use GitHub API (preferred over scraping)\n\nAlways prefer the GitHub API. It returns clean JSON — no HTML parsing needed.\n\n```\n# Repo metadata (name, description, language, stars, topics)\nGET https:\u002F\u002Fapi.github.com\u002Frepos\u002F{owner}\u002F{repo}\n\n# File tree (see what files exist — very cheap)\nGET https:\u002F\u002Fapi.github.com\u002Frepos\u002F{owner}\u002F{repo}\u002Fgit\u002Ftrees\u002F{ref}?recursive=1\n\n# Single file content (base64 encoded)\nGET https:\u002F\u002Fapi.github.com\u002Frepos\u002F{owner}\u002F{repo}\u002Fcontents\u002F{path}?ref={ref}\n\n# README only (usually enough to understand the repo)\nGET https:\u002F\u002Fapi.github.com\u002Frepos\u002F{owner}\u002F{repo}\u002Freadme\n```\n\n### Step 3 — Layered Fetch for Repos\n\n```\nLayer 1 (always do first):\n  → Fetch repo metadata + README only\n  → Can you answer the user's question now? YES → STOP. NO → continue.\n\nLayer 2 (only if needed):\n  → Fetch file tree to understand structure\n  → Identify the 1-3 most relevant files based on the question\n  → Can you answer now? YES → STOP. NO → continue.\n\nLayer 3 (last resort):\n  → Fetch specific relevant files only (never fetch all files)\n  → Prioritize: main entry point, config files, key modules\n```\n\n### Token Rules for GitHub\n\n- README alone answers ~70% of \"what does this repo do\" questions — always try it first\n- Never fetch more than 3 files in a single research turn\n- If a file exceeds ~300 lines, read only the top (imports + class\u002Ffunction signatures)\n- Decode base64 content from API before passing to context\n\n---\n\n## URL Protocol\n\nUse when the user gives a specific non-GitHub URL (docs, articles, blogs, etc.)\n\n### Step 1 — Assess the URL type\n\n| Site type | Likely works with | Notes |\n|---|---|---|\n| Static docs \u002F MDN \u002F ReadTheDocs | `read_url_content` | Fast, clean, cheap |\n| News articles \u002F blogs | `read_url_content` | Usually fine |\n| SPAs \u002F React\u002FNext.js apps | `browser_subagent` | JS-rendered |\n| Auth-gated pages | `browser_subagent` | Needs login |\n| Raw GitHub files (raw.githubusercontent) | `read_url_content` | Direct text |\n\n### Step 2 — Layered Fetch\n\n```\nLayer 1 — Skim\n  → Fetch the URL with read_url_content\n  → Read only headings (H1, H2, H3) and first paragraph\n  → Does this page contain what the user needs? NO → try a different URL or search. YES → continue.\n\nLayer 2 — Targeted Extract\n  → If the page has anchor links (e.g. \u002Fdocs\u002Fpage#section), fetch with the anchor\n  → Extract only the relevant section (200–500 tokens max)\n  → Can you answer? YES → STOP.\n\nLayer 3 — Full Fetch\n  → Fetch full page, strip boilerplate (nav, footer, ads, cookie banners, sidebars)\n  → Cap at 2000 tokens. Summarize before passing to answer.\n\nLayer 4 — Browser Subagent (last resort only)\n  → Use ONLY if read_url_content returns empty, garbled, or JS-placeholder content\n  → Instruct subagent: \"Navigate to [URL], wait for content to load, extract [specific section]\"\n  → Do NOT use browser_subagent for static pages — it's expensive\n```\n\n### What to Strip from Fetched Pages\n\nAlways remove before using fetched content:\n- Navigation menus and breadcrumbs\n- Cookie banners and GDPR notices\n- \"Related articles\" \u002F \"You might also like\" blocks\n- Footer content (copyright, links)\n- Social share buttons\n- Ads and sponsored content\n\nExtract and keep:\n- Main article \u002F documentation body\n- Code blocks\n- Tables with data\n- Numbered steps or procedures\n\n---\n\n## Search Protocol\n\nUse when the user gives a topic, question, or query — not a specific URL.\n\n### Step 1 — Sharpen the Query Before Searching\n\nDo NOT search the raw user query. Transform it first:\n\n```\nRaw: \"how to deploy fastapi on aws\"\nSharpened: \"fastapi AWS deployment tutorial 2024\"\n\nRaw: \"python async vs threads\"\nSharpened: \"Python asyncio vs threading performance comparison\"\n\nRaw: \"best way to structure react project\"\nSharpened: \"React project folder structure best practices\"\n```\n\n**Query sharpening rules:**\n- Add specificity: version numbers, technology names, \"tutorial\" \u002F \"guide\" \u002F \"comparison\"\n- Add recency if relevant: current year\n- Remove filler words: \"how do I\", \"what is the\", \"can you explain\"\n- For code questions: add the language + framework name explicitly\n\n### Step 2 — Search and Select\n\n```\n1. Run search_web with the sharpened query\n2. Get results (titles + snippets)\n3. Scan titles + snippets ONLY — do not fetch yet\n4. Pick the TOP 1-2 most relevant results (max 3 in complex cases)\n5. Skip results from: forums (if docs exist), aggregator blogs, paywalled sites\n6. Prefer: official docs, GitHub repos, well-known tech blogs, academic sources\n```\n\n### Step 3 — Fetch Selected Results\n\nApply the URL Protocol (above) to each selected URL.\nProcess results one at a time — only fetch the second URL if the first didn't answer the question.\n\n### Token Rules for Search\n\n- Never read more than 3 URLs per search query\n- If the snippet already contains the answer → do NOT fetch the full page, use the snippet\n- For factual questions (dates, names, simple facts) → snippet is usually enough\n- For procedural questions (how to do X) → fetch 1 relevant page, targeted section only\n\n---\n\n## Multi-URL Protocol\n\nUse when the user provides a list of URLs to compare or summarize.\n\n```\n1. Skim all URLs first (Layer 1 fetch for each)\n2. Group by relevance to the user's question\n3. Deep-fetch only the most relevant 1-3 URLs\n4. Summarize each in 3-5 sentences before combining\n5. Never dump raw content from multiple pages — always summarize per-source first\n```\n\n---\n\n## File Protocol\n\nUse when URL points directly to a file (PDF, .txt, .md, .csv, etc.)\n\n- `.md` \u002F `.txt` \u002F `.csv` → `read_url_content` works directly, read full content\n- `.pdf` → Use browser_subagent or a PDF extraction tool; extract text only\n- `.json` \u002F `.yaml` → `read_url_content`, parse structure, summarize schema + key values\n- Large files (>500 lines) → Read first 100 lines + last 20 lines + search for relevant sections\n\n---\n\n## Anti-Patterns (Never Do These)\n\n| Anti-pattern | Why it's bad | Do this instead |\n|---|---|---|\n| Fetching full page for a simple fact | Wastes 1000s of tokens | Use snippet or targeted anchor |\n| Using browser_subagent for static sites | Very expensive | Use read_url_content first |\n| Searching with the raw user query | Vague results | Sharpen query first |\n| Fetching 5+ search results | Token explosion | Max 3, stop when answered |\n| Dumping raw HTML into context | Noisy, wasteful | Always strip to Markdown |\n| Fetching \"just in case\" | Unnecessary tokens | Only fetch what's needed to answer |\n| Re-fetching the same URL | Redundant | Cache result in context, reuse |\n| Fetching entire GitHub repo | Extremely wasteful | README + targeted files only |\n\n---\n\n## Decision Flowchart (Quick Reference)\n\n```\nInput received\n│\n├─ GitHub URL?\n│   ├─ Fetch README + metadata via API\n│   ├─ Answered? → STOP\n│   ├─ Need more? → Fetch file tree, pick 1-3 files\n│   └─ Still need more? → Fetch specific files only\n│\n├─ Specific URL?\n│   ├─ Try read_url_content → skim headings\n│   ├─ Answered? → STOP\n│   ├─ Need more? → Targeted section fetch\n│   ├─ Still need more? → Full fetch, stripped\n│   └─ JS-rendered \u002F broken? → browser_subagent (last resort)\n│\n├─ Topic\u002Fquery?\n│   ├─ Sharpen query\n│   ├─ search_web → scan snippets\n│   ├─ Snippet enough? → Answer from snippet, STOP\n│   ├─ Need more? → Fetch top 1 result (targeted)\n│   └─ Still need more? → Fetch top 2nd result (targeted)\n│\n└─ List of URLs?\n    ├─ Skim all (Layer 1 each)\n    ├─ Deep fetch top 1-3 relevant ones\n    └─ Summarize per-source, then combine\n```\n\n---\n\n## Output Format Rules\n\nAfter fetching, structure your response as:\n\n```\nSource: [URL or \"Web search for: query\"]\nSummary: [2-5 sentences of what was found]\nAnswer: [Direct answer to user's question]\nConfidence: [High \u002F Medium \u002F Low — based on source quality]\n```\n\nFor multiple sources:\n```\nSource 1: ...\nSource 2: ...\nCombined Answer: ...\n```\n\nNever output:\n- Raw HTML fragments\n- Full page dumps\n- Unattributed information\n- More than needed to answer the question\n\n---\n\n## Token Budget Guide\n\n| Operation | Approximate token cost | When to use |\n|---|---|---|\n| GitHub README fetch | ~300–800 tokens | Always first for repos |\n| GitHub API metadata | ~200 tokens | Always for repos |\n| Skim (headings only) | ~100–200 tokens | Always first for URLs |\n| Targeted section fetch | ~300–600 tokens | When skim isn't enough |\n| Full page fetch (stripped) | ~1000–2000 tokens | Only when targeted fails |\n| browser_subagent | ~2000–5000 tokens | Last resort only |\n| Search snippet scan | ~300–500 tokens | Always before fetching |\n\n**Rule of thumb:** If you're about to spend >2000 tokens on a fetch, ask yourself if there's a cheaper path first.\n\n---\n\n## Limitations\n\n- **JavaScript Reliance**: Standard fetching may not fully render Single Page Applications (SPAs). You must fallback to `browser_subagent` for these, which is slower and more expensive.\n- **Paywalls & Protections**: This skill cannot bypass CAPTCHAs, bot protections (e.g., strict Cloudflare rules), or hard paywalls.\n- **GitHub API Limits**: Frequent GitHub API requests without authentication may hit rate limits.\n","# Efficient Web Research Skill\n\nA protocol for accessing web content in the most token-efficient, accurate, and structured way —\nusing the right tool at the right depth, and stopping as soon as the question is answerable.\n\n---\n\n## When to Use\n- Use this skill when the task matches this description: Protocol for token-efficient web research. Use when accessing URLs, GitHub repos, or running search queries. Prevents full-page fetching waste.\n\n## Core Principle\n\n> **Fetch the minimum needed to answer. Skim before you dive. Stop when you can answer.**\n\nEvery unnecessary fetch wastes tokens and adds noise. This skill enforces a layered approach\nwhere you escalate fetch depth only when shallower layers fail.\n\n---\n\n## Step 1 — Classify the Input\n\nBefore fetching anything, identify what kind of input you received:\n\n| Input Type | Example | Go To |\n|---|---|---|\n| GitHub repo URL | `github.com\u002Fuser\u002Frepo` | [GitHub Protocol](#github-protocol) |\n| Specific page URL | `docs.python.org\u002F3\u002Flibrary\u002Fos` | [URL Protocol](#url-protocol) |\n| Topic \u002F query (no URL) | \"how does RAFT consensus work\" | [Search Protocol](#search-protocol) |\n| Multiple URLs | List of links | [Multi-URL Protocol](#multi-url-protocol) |\n| PDF \u002F file link | `.pdf`, `.txt`, `.md` URL | [File Protocol](#file-protocol) |\n\n---\n\n## GitHub Protocol\n\nUse when input is a GitHub URL (repo, file, PR, issue, etc.)\n\n### Step 1 — Parse the URL\n\n```\ngithub.com\u002F{owner}\u002F{repo}                → Repo root\ngithub.com\u002F{owner}\u002F{repo}\u002Ftree\u002F{branch}  → Directory\ngithub.com\u002F{owner}\u002F{repo}\u002Fblob\u002F{branch}\u002F{path} → Single file\ngithub.com\u002F{owner}\u002F{repo}\u002Fissues\u002F{n}     → Issue\ngithub.com\u002F{owner}\u002F{repo}\u002Fpull\u002F{n}       → Pull request\n```\n\n### Step 2 — Use GitHub API (preferred over scraping)\n\nAlways prefer the GitHub API. It returns clean JSON — no HTML parsing needed.\n\n```\n# Repo metadata (name, description, language, stars, topics)\nGET https:\u002F\u002Fapi.github.com\u002Frepos\u002F{owner}\u002F{repo}\n\n# File tree (see what files exist — very cheap)\nGET https:\u002F\u002Fapi.github.com\u002Frepos\u002F{owner}\u002F{repo}\u002Fgit\u002Ftrees\u002F{ref}?recursive=1\n\n# Single file content (base64 encoded)\nGET https:\u002F\u002Fapi.github.com\u002Frepos\u002F{owner}\u002F{repo}\u002Fcontents\u002F{path}?ref={ref}\n\n# README only (usually enough to understand the repo)\nGET https:\u002F\u002Fapi.github.com\u002Frepos\u002F{owner}\u002F{repo}\u002Freadme\n```\n\n### Step 3 — Layered Fetch for Repos\n\n```\nLayer 1 (always do first):\n  → Fetch repo metadata + README only\n  → Can you answer the user's question now? YES → STOP. NO → continue.\n\nLayer 2 (only if needed):\n  → Fetch file tree to understand structure\n  → Identify the 1-3 most relevant files based on the question\n  → Can you answer now? YES → STOP. NO → continue.\n\nLayer 3 (last resort):\n  → Fetch specific relevant files only (never fetch all files)\n  → Prioritize: main entry point, config files, key modules\n```\n\n### Token Rules for GitHub\n\n- README alone answers ~70% of \"what does this repo do\" questions — always try it first\n- Never fetch more than 3 files in a single research turn\n- If a file exceeds ~300 lines, read only the top (imports + class\u002Ffunction signatures)\n- Decode base64 content from API before passing to context\n\n---\n\n## URL Protocol\n\nUse when the user gives a specific non-GitHub URL (docs, articles, blogs, etc.)\n\n### Step 1 — Assess the URL type\n\n| Site type | Likely works with | Notes |\n|---|---|---|\n| Static docs \u002F MDN \u002F ReadTheDocs | `read_url_content` | Fast, clean, cheap |\n| News articles \u002F blogs | `read_url_content` | Usually fine |\n| SPAs \u002F React\u002FNext.js apps | `browser_subagent` | JS-rendered |\n| Auth-gated pages | `browser_subagent` | Needs login |\n| Raw GitHub files (raw.githubusercontent) | `read_url_content` | Direct text |\n\n### Step 2 — Layered Fetch\n\n```\nLayer 1 — Skim\n  → Fetch the URL with read_url_content\n  → Read only headings (H1, H2, H3) and first paragraph\n  → Does this page contain what the user needs? NO → try a different URL or search. YES → continue.\n\nLayer 2 — Targeted Extract\n  → If the page has anchor links (e.g. \u002Fdocs\u002Fpage#section), fetch with the anchor\n  → Extract only the relevant section (200–500 tokens max)\n  → Can you answer? YES → STOP.\n\nLayer 3 — Full Fetch\n  → Fetch full page, strip boilerplate (nav, footer, ads, cookie banners, sidebars)\n  → Cap at 2000 tokens. Summarize before passing to answer.\n\nLayer 4 — Browser Subagent (last resort only)\n  → Use ONLY if read_url_content returns empty, garbled, or JS-placeholder content\n  → Instruct subagent: \"Navigate to [URL], wait for content to load, extract [specific section]\"\n  → Do NOT use browser_subagent for static pages — it's expensive\n```\n\n### What to Strip from Fetched Pages\n\nAlways remove before using fetched content:\n- Navigation menus and breadcrumbs\n- Cookie banners and GDPR notices\n- \"Related articles\" \u002F \"You might also like\" blocks\n- Footer content (copyright, links)\n- Social share buttons\n- Ads and sponsored content\n\nExtract and keep:\n- Main article \u002F documentation body\n- Code blocks\n- Tables with data\n- Numbered steps or procedures\n\n---\n\n## Search Protocol\n\nUse when the user gives a topic, question, or query — not a specific URL.\n\n### Step 1 — Sharpen the Query Before Searching\n\nDo NOT search the raw user query. Transform it first:\n\n```\nRaw: \"how to deploy fastapi on aws\"\nSharpened: \"fastapi AWS deployment tutorial 2024\"\n\nRaw: \"python async vs threads\"\nSharpened: \"Python asyncio vs threading performance comparison\"\n\nRaw: \"best way to structure react project\"\nSharpened: \"React project folder structure best practices\"\n```\n\n**Query sharpening rules:**\n- Add specificity: version numbers, technology names, \"tutorial\" \u002F \"guide\" \u002F \"comparison\"\n- Add recency if relevant: current year\n- Remove filler words: \"how do I\", \"what is the\", \"can you explain\"\n- For code questions: add the language + framework name explicitly\n\n### Step 2 — Search and Select\n\n```\n1. Run search_web with the sharpened query\n2. Get results (titles + snippets)\n3. Scan titles + snippets ONLY — do not fetch yet\n4. Pick the TOP 1-2 most relevant results (max 3 in complex cases)\n5. Skip results from: forums (if docs exist), aggregator blogs, paywalled sites\n6. Prefer: official docs, GitHub repos, well-known tech blogs, academic sources\n```\n\n### Step 3 — Fetch Selected Results\n\nApply the URL Protocol (above) to each selected URL.\nProcess results one at a time — only fetch the second URL if the first didn't answer the question.\n\n### Token Rules for Search\n\n- Never read more than 3 URLs per search query\n- If the snippet already contains the answer → do NOT fetch the full page, use the snippet\n- For factual questions (dates, names, simple facts) → snippet is usually enough\n- For procedural questions (how to do X) → fetch 1 relevant page, targeted section only\n\n---\n\n## Multi-URL Protocol\n\nUse when the user provides a list of URLs to compare or summarize.\n\n```\n1. Skim all URLs first (Layer 1 fetch for each)\n2. Group by relevance to the user's question\n3. Deep-fetch only the most relevant 1-3 URLs\n4. Summarize each in 3-5 sentences before combining\n5. Never dump raw content from multiple pages — always summarize per-source first\n```\n\n---\n\n## File Protocol\n\nUse when URL points directly to a file (PDF, .txt, .md, .csv, etc.)\n\n- `.md` \u002F `.txt` \u002F `.csv` → `read_url_content` works directly, read full content\n- `.pdf` → Use browser_subagent or a PDF extraction tool; extract text only\n- `.json` \u002F `.yaml` → `read_url_content`, parse structure, summarize schema + key values\n- Large files (>500 lines) → Read first 100 lines + last 20 lines + search for relevant sections\n\n---\n\n## Anti-Patterns (Never Do These)\n\n| Anti-pattern | Why it's bad | Do this instead |\n|---|---|---|\n| Fetching full page for a simple fact | Wastes 1000s of tokens | Use snippet or targeted anchor |\n| Using browser_subagent for static sites | Very expensive | Use read_url_content first |\n| Searching with the raw user query | Vague results | Sharpen query first |\n| Fetching 5+ search results | Token explosion | Max 3, stop when answered |\n| Dumping raw HTML into context | Noisy, wasteful | Always strip to Markdown |\n| Fetching \"just in case\" | Unnecessary tokens | Only fetch what's needed to answer |\n| Re-fetching the same URL | Redundant | Cache result in context, reuse |\n| Fetching entire GitHub repo | Extremely wasteful | README + targeted files only |\n\n---\n\n## Decision Flowchart (Quick Reference)\n\n```\nInput received\n│\n├─ GitHub URL?\n│   ├─ Fetch README + metadata via API\n│   ├─ Answered? → STOP\n│   ├─ Need more? → Fetch file tree, pick 1-3 files\n│   └─ Still need more? → Fetch specific files only\n│\n├─ Specific URL?\n│   ├─ Try read_url_content → skim headings\n│   ├─ Answered? → STOP\n│   ├─ Need more? → Targeted section fetch\n│   ├─ Still need more? → Full fetch, stripped\n│   └─ JS-rendered \u002F broken? → browser_subagent (last resort)\n│\n├─ Topic\u002Fquery?\n│   ├─ Sharpen query\n│   ├─ search_web → scan snippets\n│   ├─ Snippet enough? → Answer from snippet, STOP\n│   ├─ Need more? → Fetch top 1 result (targeted)\n│   └─ Still need more? → Fetch top 2nd result (targeted)\n│\n└─ List of URLs?\n    ├─ Skim all (Layer 1 each)\n    ├─ Deep fetch top 1-3 relevant ones\n    └─ Summarize per-source, then combine\n```\n\n---\n\n## Output Format Rules\n\nAfter fetching, structure your response as:\n\n```\nSource: [URL or \"Web search for: query\"]\nSummary: [2-5 sentences of what was found]\nAnswer: [Direct answer to user's question]\nConfidence: [High \u002F Medium \u002F Low — based on source quality]\n```\n\nFor multiple sources:\n```\nSource 1: ...\nSource 2: ...\nCombined Answer: ...\n```\n\nNever output:\n- Raw HTML fragments\n- Full page dumps\n- Unattributed information\n- More than needed to answer the question\n\n---\n\n## Token Budget Guide\n\n| Operation | Approximate token cost | When to use |\n|---|---|---|\n| GitHub README fetch | ~300–800 tokens | Always first for repos |\n| GitHub API metadata | ~200 tokens | Always for repos |\n| Skim (headings only) | ~100–200 tokens | Always first for URLs |\n| Targeted section fetch | ~300–600 tokens | When skim isn't enough |\n| Full page fetch (stripped) | ~1000–2000 tokens | Only when targeted fails |\n| browser_subagent | ~2000–5000 tokens | Last resort only |\n| Search snippet scan | ~300–500 tokens | Always before fetching |\n\n**Rule of thumb:** If you're about to spend >2000 tokens on a fetch, ask yourself if there's a cheaper path first.\n\n---\n\n## Limitations\n\n- **JavaScript Reliance**: Standard fetching may not fully render Single Page Applications (SPAs). You must fallback to `browser_subagent` for these, which is slower and more expensive.\n- **Paywalls & Protections**: This skill cannot bypass CAPTCHAs, bot protections (e.g., strict Cloudflare rules), or hard paywalls.\n- **GitHub API Limits**: Frequent GitHub API requests without authentication may hit rate limits.\n",{"name":5,"risk":24,"source":25,"category":26,"date_added":27,"description":7},"safe","community","development","2026-09-04","main","2d1750717747097b602acf30bd09a7cbd4b4c6ea","awesome_skills\u002Fdevelopment\u002Fefficient-web-research","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous","2026-11-09 11:45:36",{"source":8,"skillName":5,"commands":35,"prompt":45,"zipAvailable":46},[36,39,42],{"tool":37,"command":38},"npx","npx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill efficient-web-research",{"tool":40,"command":41},"bunx","bunx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill efficient-web-research",{"tool":43,"command":44},"pnpm","pnpm dlx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill efficient-web-research","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fdevelopment\u002Fefficient-web-research\u002FSKILL.md",true,[],[49,57,64,71,78,85],{"skillNo":50,"name":51,"nameZh":6,"description":52,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":53,"categories":54,"keywords":55,"publishedAt":56},"01M7FFFZPC2025JQQHPHDX154J","dei-assessment","Assess diversity, equity, and inclusion with metrics. TRIGGERS - Use when user needs help with dei-assessment related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fdei-assessment\u002FSKILL.md",[],[],"2026-11-16 10:05:34",{"skillNo":58,"name":59,"nameZh":6,"description":60,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":61,"categories":62,"keywords":63,"publishedAt":56},"01M7FFFWN8H4WA7A6Y398QKB6J","dei-action-plan","Create DEI action plans with goals and initiatives. TRIGGERS - Use when user needs help with dei-action-plan related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fdei-action-plan\u002FSKILL.md",[],[],{"skillNo":65,"name":66,"nameZh":6,"description":67,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":68,"categories":69,"keywords":70,"publishedAt":56},"01M7FFFSHTM0QZRVZ2NRJTJ0PY","culture-fit-assessment","Design culture add assessment processes. TRIGGERS - Use when user needs help with culture-fit-assessment related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fculture-fit-assessment\u002FSKILL.md",[],[],{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":56},"01M7FFFMA12RKEQA38KHJSXXPS","cowork-to-code-bridge","Use an already-installed, independently verified cowork-to-code bridge to run narrowly approved actions on the user's own macOS, Linux, or WSL2 machine through a local file queue.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcowork-to-code-bridge\u002FSKILL.md",[],[],{"skillNo":79,"name":80,"nameZh":6,"description":81,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":56},"01M7FFFH1F6M8YRGW5V487GH1K","counter-offer-response","Draft counter-offer responses for retention. TRIGGERS - Use when user needs help with counter-offer-response related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcounter-offer-response\u002FSKILL.md",[],[],{"skillNo":86,"name":87,"nameZh":6,"description":88,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":89,"categories":90,"keywords":91,"publishedAt":56},"01M7FFFDGEHV8Y7J9N40HZXW52","continuous-feedback-system","Design continuous feedback systems. TRIGGERS - Use when user needs help with continuous-feedback-system related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcontinuous-feedback-system\u002FSKILL.md",[],[]]