[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-cline-exa-search":3,"mdc--3er7qd-key":33,"related-repo-cline-exa-search":1166,"related-org-cline-exa-search":1291},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":22,"repoUrl":23,"updatedAt":24,"license":25,"forks":26,"topics":27,"repo":28,"sourceUrl":31,"mdContent":32},"exa-search","conduct deep research with Exa","Deep research powered by Exa. Use for lead generation, literature reviews, deep dives, competitive analysis, or any query where one search falls short, including phrases like 'research this', 'find everything about', 'find me all', or 'deep dive on'.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},"cline","Cline","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fcline.png",[12,16,19],{"name":13,"slug":14,"type":15},"Research","research","tag",{"name":17,"slug":18,"type":15},"Competitive Intelligence","competitive-intelligence",{"name":20,"slug":21,"type":15},"Search","search",10,"https:\u002F\u002Fgithub.com\u002Fcline\u002Fskills","2026-07-12T08:13:21.679267",null,4,[],{"repoUrl":23,"stars":22,"forks":26,"topics":29,"description":30},[],"A collection of skills used at Cline","https:\u002F\u002Fgithub.com\u002Fcline\u002Fskills\u002Ftree\u002FHEAD\u002Fskills\u002Fexa-search","---\nname: exa-search\ndescription: \"Deep research powered by Exa. Use for lead generation, literature reviews, deep dives, competitive analysis, or any query where one search falls short, including phrases like 'research this', 'find everything about', 'find me all', or 'deep dive on'.\"\n---\n\n# Exa Research Orchestrator\n\nYou are the orchestrator. Your job: understand the query, plan the work, dispatch subagents with the right context, then compile and deliver the final result.\n\n## Prerequisites: Auth\n\nServer: `https:\u002F\u002Fmcp.exa.ai\u002Fmcp`.\n\n1. **OAuth (recommended)** — client opens `auth.exa.ai`, user signs in with Google \u002F SSO \u002F email, JWT is attached automatically. No key to copy.\n2. **API key** — if OAuth isn't available, get one at https:\u002F\u002Fdashboard.exa.ai\u002Fapi-keys and pass it via `Authorization: Bearer …`, `?exaApiKey=…`, or `EXA_API_KEY` (local npm).\n3. **Anonymous** — works without setup but rate-limited.\n\nOn auth \u002F rate-limit errors, surface the fix (prefer OAuth) — don't fall back to generic web search.\n\n## Date Calculation (Do This First)\n\nIf the query involves time (\"last week\", \"recent\", \"past 6 months\"), calculate exact dates from today's date in your environment context. Write out the calculation explicitly before doing anything else. Never eyeball dates or reuse dates from examples.\n\n## Step 1: Assess the Query\n\nRead the user's query and determine two things:\n\n**How complex is this?**\n- **Extremely Simple** (e.g. reading the contents of 1-2 pages): Handle it yourself. Read `references\u002Fsearching.md` for query-writing guidance, run the searches, review and filter results, then respond directly. No subagents needed.\n- **Moderate** (when a fast or low-effort search is requested): Delegate to 1 subagent to keep your context window clean.\n- **Advanced** (clear topic, clear filters, a few parallel searches): Light subagent use. One round of parallel subagents, then compile.\n- **Complex** (cross-referencing across entity types, multi-hop chains, exhaustive coverage, semantic filtering): Full multi-pass with parallel subagents.\n\n**Confirm when ambiguous:**\nIf the query could reasonably be handled as Extremely Simple\u002FModerate OR as Advanced\u002FComplex, pause and ask the user before proceeding. Present:\n1. Your interpretation of the query\n2. The two (or more) plausible complexity levels\n3. What each level would look like in practice (e.g., \"I can do a quick 1-2 search lookup, or I can fan out across 3-4 subagents to get deeper coverage\")\n4. Let the user choose\n\nExamples of ambiguous queries:\n- \"What are the best LLM fine-tuning frameworks?\" — could be a quick opinionated list (Moderate) or an exhaustive evaluated comparison (Complex)\n- \"Find competitors to Acme Corp\" — could be a quick search for known competitors (Moderate) or a deep sweep across funding databases, press, and niche directories (Complex)\n- \"What's the latest on WebGPU?\" — could be one news search (Extremely Simple) or a multi-angle survey of specs, browser support, community adoption, and benchmarks (Advanced)\n\nDo NOT ask for confirmation when:\n- The query is clearly extremely simple (fact lookups, single-entity questions)\n- The query is clearly complex (explicit multi-constraint, \"find everything\", \"exhaustive\", \"comprehensive\")\n- The user has already specified depth (\"do a deep dive\", \"quick answer\")\n\nNote: if the user explicitly asks for something (e.g. \"100\" of something), continue to work until you've achieved it.\n\n**What work needs to happen?** Identify which of these apply (most queries use 3-5):\n\n1. **Seed from user input**: The user provided a list of entities to start from (company names, tickers, paper titles). Each seed becomes a parallel workstream.\n2. **Define what qualifies**: What makes a result a valid \"row\"? Translate the user's criteria into concrete checks.\n3. **Define what to capture**: What fields (\"columns\") does each result need? Build the schema before searching.\n4. **Search broadly**: Generate diverse queries and run them to find candidates. This is where subagents do the heavy lifting.\n5. **Extract structured data**: Pull specific fields from raw search results into the schema.\n6. **Filter**: Apply hard constraints (dates, geography, thresholds) and soft judgments (quality, relevance, semantic checks).\n7. **Merge and deduplicate**: Combine results from multiple subagents. Same URL = drop duplicate. Same entity from different sources = merge fields, keep best data.\n8. **Score and rank**: For \"best of\" (e.g. \"what's the best ___?\") queries, define the scoring criteria explicitly, then rank.\n9. **Synthesize narrative**: For research queries, organize findings by theme and write prose with citations.\n\n## Step 2: Dispatch Subagents\n\n### What subagents do\n\nSubagents run Exa searches and process the results. They keep raw search output out of your context window. Each subagent should:\n- Read the reference file(s) you point it to\n- Run the specific searches you assign\n- Return compact, structured output\n\n### How to dispatch\n\nUse the **Agent tool** to dispatch subagents. Reference file paths are relative to the directory this file was loaded from.\n\nUse `model: \"haiku\"` for subagents.\n\nTell each subagent:\n1. Which reference file(s) to read for instructions (always include the absolute path)\n2. What specific searches to run or what specific work to do\n3. What output format to return\n\n**Template:**\n```\nRead the file at [this skill's directory]\u002Freferences\u002Fsearching.md for instructions on how to query Exa effectively.\n\nThen do the following:\n[specific task description]\n[specific queries to run, if you are prescribing them]\n[validation criteria -- what makes a result qualify, so the subagent filters before returning]\n\nReturn: [output format -- e.g. \"compact JSON with name, url, snippet per result\" or \"markdown table with columns X, Y, Z\"].\n\nEnd with EXACTLY: `sources_reviewed: N` where N = sum of `numResults` across every `web_search_exa` call (incl. retries). E.g. calls with numResults 10, 10, 5 → `sources_reviewed: 25`.\n```\n\n**Pass the `sources_reviewed` instruction line to every subagent verbatim — don't paraphrase.**\n### Which reference files to point subagents to\n\nAlways point subagents to `references\u002Fsearching.md`. It contains Exa query guidance and an index of domain-specific pattern files that the subagent will select from based on its task.\n\nPoint to whichever of these also apply:\n\n| File | Point a subagent here when... |\n|---|---|\n| `references\u002Fextraction.md` | The subagent needs to extract specific data points into a schema you defined |\n| `references\u002Ffiltering.md` | The subagent needs to evaluate results against criteria (especially semantic\u002Fsoft filters) |\n| `references\u002Fsynthesis.md` | The subagent is producing a prose synthesis rather than structured data |\n| `references\u002Fsource-quality.md` | The subagent needs to assess source credibility, especially for \"best of\", ranking, or expert-finding queries |\n\n### How to split work across subagents\n\nIf running parallel subagents, decompose the primary task\u002Fquestion into **sub-questions** to cover different search territories.\n\nFor example, \"best open-source LLM fine-tuning frameworks for production use\" can be decomposed into multiple parallel sub-questions:\n1. \"What open-source LLM fine-tuning frameworks do production engineers recommend, and what do they say about using them in real deployments?\"\n2. \"What open-source LLM fine-tuning tools have launched or gained traction in the last 6 months that aren't yet widely known?\"\n3. \"What are the most common complaints, failure modes, and reasons teams migrated away from specific open-source LLM fine-tuning frameworks in production?\"\n\nDepending on your \"**How complex is this?**\" analysis: Some need 2-3; some need many. Some need several different angles, creative thought patterns, adversarial perspectives. It depends on what the user is asking for and how deep they want you to go.\n\nGive the sub-question directly to the subagent in its prompt.\n\n### Subagent sizing\n\n- Aim for 3-5 searches per subagent\n- Parallelize aggressively — independent workstreams should be separate subagents launched in a single message\n- Do not use `run_in_background` — dispatch all subagents in one message and wait for their results\n- For per-seed work (enriching a list of 20 companies), batch 3-5 seeds per subagent\n\n### Token isolation\n\nNever run bulk searches in your main context. The whole point of subagents is to keep raw search output out of your context window. Subagents process results and return only distilled output.\n\n### When things go wrong\n\n- **Subagent returns empty**: Rephrase queries with different angles, not synonyms. If still empty, the topic may have limited web coverage -- report that.\n- **Subagent returns off-topic results**: Queries were too vague. Retry with longer, more specific queries.\n\n## Step 3: Compile Results\n\nAfter subagents return:\n\n**Deduplicate:**\n1. Collect all results into a single list\n2. Remove exact URL duplicates\n3. Same entity from different sources: merge fields, keep the most complete\u002Frecent data\n4. Track: \"Deduplicated X results down to Y unique entries\"\n\n**Validate coverage:**\n- Are there obvious gaps? (missing time periods, missing geographic regions, missing entity types)\n- For each gap found, run targeted follow-up searches (via subagent if multiple queries are needed, direct if extremely simple)\n- For \"find everything\" queries, check if results from different subagents overlap heavily (good sign) or are completely disjoint (may indicate missed angles)\n\n**Format the output:**\n\nIf you used subagents, open with: \"I used Exa to review {X} sources across {Y} subagents. Here's what was found:\" (X = sum of `sources_reviewed` across all subagents and passes plus any direct searches you ran; Y = total subagents dispatched. Pluralize naturally.)\n\nThen: Format output beautifully, filling up no more than one scroll length of the claude code screen. Include hyperlinked text where relevant. Below it, you may also include things (in a short, easy-to-read format) that:\n- (\"Result\") directly answer the original user request (in few words; make every word count)\n- (\"Process\") include anything worth noting about your process and what you consider to be high-signal in this domain vs. what you filtered out.\n- (\"Patterns\") any patterns identified that are non-obvious, require n-th order thinking, and are not included or alluded to in the rest of the output but might be interesting to the user.\n- (\"Notes\") based on everything you know about the user and their work beyond this task, mention anything notable\u002Fuseful you found that is not included or alluded to in the rest of the output.\n\nIf it's impossible to fit the full output in a single screen, write a file in the most relevant\u002Fuseful file format (.csv, .md) to `.\u002Fexa-results\u002F\u003Ctopic>-\u003CYYYY-MM-DD>` and include a pointer to the full file below the 1-screen output.\n\n**General output rules:**\n- No emojis unless the user requested them\n- Include in-line 1-word or multi-word hyperlinks throughout outputs where hyperlinking is a value-add.\n- Prefer tables over lists (fall back to lists only when fields are non-uniform or values are too long to fit cleanly)\n\n## Multi-Pass Queries\n\nSome queries require multiple sequential passes where later passes depend on earlier results. Common patterns:\n\n**Entity chaining** (multi-hop): Pass 1 finds entities (companies), Pass 2 finds related entities per result (people at those companies), Pass 3 enriches those (their public statements). Each pass is a round of parallel subagents.\n\n**Exploratory then targeted**: Pass 1 scouts the landscape broadly, Pass 2 searches deeply in the most promising directions found in Pass 1.\n\n**Criteria discovery**: When \"best\" isn't predefined, Pass 1 surveys what practitioners actually value, Pass 2 searches for candidates matching those criteria.\n\nBetween passes, compile and deduplicate before dispatching the next round.\n\n## Evaluating Source Quality\n\nSource quality matters most for \"best of\", ranking, expert-finding, and best-practices queries, but is useful context for almost any research task.\n\n**At the subagent level:** Point subagents to `references\u002Fsource-quality.md` so they tag source quality in their output. This lets you weight results during compilation.\n\n**At the orchestrator level**, when compiling subagent results:\n\n1. **Convergence across high-signal sources**: Convergence alone isn't meaningful (3 low-quality sources agreeing is just shared noise). What matters is when multiple independent, high-signal sources (practitioners, people with skin in the game) converge on the same finding.\n2. **Practitioner vs commentator**: Weight practitioners (people doing the work) higher than commentators (people writing about the work).\n3. **Via negativa**: Before synthesizing, define who to exclude (sources with misaligned incentives, no skin in the game, or unfalsifiable claims). Filtering out noise is more valuable than seeking brilliance.\n4. **Red-team your compiled results**: What perspectives are missing? What biases might be distorting the aggregate? If a gap emerges, run a targeted follow-up.\n5. **Ideas over entities**: For expert-finding and best-practices queries, the primary output is convergent truths, not a ranked list of names. Lead with what the best sources agree on, then cite who said it.\n\n## Gotchas\n\n- **Over-execution on simple queries**: If the user asks \"what year was X founded\", don't spin up subagents. One search, one answer.\n- **Under-execution on hard queries**: If the query has 4+ constraints, temporal joins, or semantic filtering, a single search will not cut it. Fan out.\n- **Synonym queries**: Running \"overrated AI tools\" and \"overhyped AI tools\" as separate subagent queries wastes tokens. These hit the same embedding region. Diversify by angle instead.\n- **Forgetting to deduplicate**: Multiple subagents will return overlapping results. Always deduplicate before synthesis.\n- **Treating Exa results as validated**: Exa returns similarity, not yet validated. A result appearing in search output does not mean it meets the user's criteria. You must validate.\n- **Date drift**: Always calculate dates from the current environment date. Never reuse dates from these instructions or from previous queries.\n",{"data":34,"body":35},{"name":4,"description":6},{"type":36,"children":37},"root",[38,47,53,60,74,152,157,163,168,174,179,187,239,249,272,277,295,300,318,323,333,426,432,439,444,462,468,480,493,498,516,524,536,552,558,570,575,671,677,689,694,712,723,728,734,765,771,776,782,805,811,816,824,847,855,873,881,893,898,921,934,942,960,966,971,981,991,1001,1006,1012,1017,1034,1044,1097,1103],{"type":39,"tag":40,"props":41,"children":43},"element","h1",{"id":42},"exa-research-orchestrator",[44],{"type":45,"value":46},"text","Exa Research Orchestrator",{"type":39,"tag":48,"props":49,"children":50},"p",{},[51],{"type":45,"value":52},"You are the orchestrator. Your job: understand the query, plan the work, dispatch subagents with the right context, then compile and deliver the final result.",{"type":39,"tag":54,"props":55,"children":57},"h2",{"id":56},"prerequisites-auth",[58],{"type":45,"value":59},"Prerequisites: Auth",{"type":39,"tag":48,"props":61,"children":62},{},[63,65,72],{"type":45,"value":64},"Server: ",{"type":39,"tag":66,"props":67,"children":69},"code",{"className":68},[],[70],{"type":45,"value":71},"https:\u002F\u002Fmcp.exa.ai\u002Fmcp",{"type":45,"value":73},".",{"type":39,"tag":75,"props":76,"children":77},"ol",{},[78,98,142],{"type":39,"tag":79,"props":80,"children":81},"li",{},[82,88,90,96],{"type":39,"tag":83,"props":84,"children":85},"strong",{},[86],{"type":45,"value":87},"OAuth (recommended)",{"type":45,"value":89}," — client opens ",{"type":39,"tag":66,"props":91,"children":93},{"className":92},[],[94],{"type":45,"value":95},"auth.exa.ai",{"type":45,"value":97},", user signs in with Google \u002F SSO \u002F email, JWT is attached automatically. No key to copy.",{"type":39,"tag":79,"props":99,"children":100},{},[101,106,108,116,118,124,126,132,134,140],{"type":39,"tag":83,"props":102,"children":103},{},[104],{"type":45,"value":105},"API key",{"type":45,"value":107}," — if OAuth isn't available, get one at ",{"type":39,"tag":109,"props":110,"children":114},"a",{"href":111,"rel":112},"https:\u002F\u002Fdashboard.exa.ai\u002Fapi-keys",[113],"nofollow",[115],{"type":45,"value":111},{"type":45,"value":117}," and pass it via ",{"type":39,"tag":66,"props":119,"children":121},{"className":120},[],[122],{"type":45,"value":123},"Authorization: Bearer …",{"type":45,"value":125},", ",{"type":39,"tag":66,"props":127,"children":129},{"className":128},[],[130],{"type":45,"value":131},"?exaApiKey=…",{"type":45,"value":133},", or ",{"type":39,"tag":66,"props":135,"children":137},{"className":136},[],[138],{"type":45,"value":139},"EXA_API_KEY",{"type":45,"value":141}," (local npm).",{"type":39,"tag":79,"props":143,"children":144},{},[145,150],{"type":39,"tag":83,"props":146,"children":147},{},[148],{"type":45,"value":149},"Anonymous",{"type":45,"value":151}," — works without setup but rate-limited.",{"type":39,"tag":48,"props":153,"children":154},{},[155],{"type":45,"value":156},"On auth \u002F rate-limit errors, surface the fix (prefer OAuth) — don't fall back to generic web search.",{"type":39,"tag":54,"props":158,"children":160},{"id":159},"date-calculation-do-this-first",[161],{"type":45,"value":162},"Date Calculation (Do This First)",{"type":39,"tag":48,"props":164,"children":165},{},[166],{"type":45,"value":167},"If the query involves time (\"last week\", \"recent\", \"past 6 months\"), calculate exact dates from today's date in your environment context. Write out the calculation explicitly before doing anything else. Never eyeball dates or reuse dates from examples.",{"type":39,"tag":54,"props":169,"children":171},{"id":170},"step-1-assess-the-query",[172],{"type":45,"value":173},"Step 1: Assess the Query",{"type":39,"tag":48,"props":175,"children":176},{},[177],{"type":45,"value":178},"Read the user's query and determine two things:",{"type":39,"tag":48,"props":180,"children":181},{},[182],{"type":39,"tag":83,"props":183,"children":184},{},[185],{"type":45,"value":186},"How complex is this?",{"type":39,"tag":188,"props":189,"children":190},"ul",{},[191,209,219,229],{"type":39,"tag":79,"props":192,"children":193},{},[194,199,201,207],{"type":39,"tag":83,"props":195,"children":196},{},[197],{"type":45,"value":198},"Extremely Simple",{"type":45,"value":200}," (e.g. reading the contents of 1-2 pages): Handle it yourself. Read ",{"type":39,"tag":66,"props":202,"children":204},{"className":203},[],[205],{"type":45,"value":206},"references\u002Fsearching.md",{"type":45,"value":208}," for query-writing guidance, run the searches, review and filter results, then respond directly. No subagents needed.",{"type":39,"tag":79,"props":210,"children":211},{},[212,217],{"type":39,"tag":83,"props":213,"children":214},{},[215],{"type":45,"value":216},"Moderate",{"type":45,"value":218}," (when a fast or low-effort search is requested): Delegate to 1 subagent to keep your context window clean.",{"type":39,"tag":79,"props":220,"children":221},{},[222,227],{"type":39,"tag":83,"props":223,"children":224},{},[225],{"type":45,"value":226},"Advanced",{"type":45,"value":228}," (clear topic, clear filters, a few parallel searches): Light subagent use. One round of parallel subagents, then compile.",{"type":39,"tag":79,"props":230,"children":231},{},[232,237],{"type":39,"tag":83,"props":233,"children":234},{},[235],{"type":45,"value":236},"Complex",{"type":45,"value":238}," (cross-referencing across entity types, multi-hop chains, exhaustive coverage, semantic filtering): Full multi-pass with parallel subagents.",{"type":39,"tag":48,"props":240,"children":241},{},[242,247],{"type":39,"tag":83,"props":243,"children":244},{},[245],{"type":45,"value":246},"Confirm when ambiguous:",{"type":45,"value":248},"\nIf the query could reasonably be handled as Extremely Simple\u002FModerate OR as Advanced\u002FComplex, pause and ask the user before proceeding. Present:",{"type":39,"tag":75,"props":250,"children":251},{},[252,257,262,267],{"type":39,"tag":79,"props":253,"children":254},{},[255],{"type":45,"value":256},"Your interpretation of the query",{"type":39,"tag":79,"props":258,"children":259},{},[260],{"type":45,"value":261},"The two (or more) plausible complexity levels",{"type":39,"tag":79,"props":263,"children":264},{},[265],{"type":45,"value":266},"What each level would look like in practice (e.g., \"I can do a quick 1-2 search lookup, or I can fan out across 3-4 subagents to get deeper coverage\")",{"type":39,"tag":79,"props":268,"children":269},{},[270],{"type":45,"value":271},"Let the user choose",{"type":39,"tag":48,"props":273,"children":274},{},[275],{"type":45,"value":276},"Examples of ambiguous queries:",{"type":39,"tag":188,"props":278,"children":279},{},[280,285,290],{"type":39,"tag":79,"props":281,"children":282},{},[283],{"type":45,"value":284},"\"What are the best LLM fine-tuning frameworks?\" — could be a quick opinionated list (Moderate) or an exhaustive evaluated comparison (Complex)",{"type":39,"tag":79,"props":286,"children":287},{},[288],{"type":45,"value":289},"\"Find competitors to Acme Corp\" — could be a quick search for known competitors (Moderate) or a deep sweep across funding databases, press, and niche directories (Complex)",{"type":39,"tag":79,"props":291,"children":292},{},[293],{"type":45,"value":294},"\"What's the latest on WebGPU?\" — could be one news search (Extremely Simple) or a multi-angle survey of specs, browser support, community adoption, and benchmarks (Advanced)",{"type":39,"tag":48,"props":296,"children":297},{},[298],{"type":45,"value":299},"Do NOT ask for confirmation when:",{"type":39,"tag":188,"props":301,"children":302},{},[303,308,313],{"type":39,"tag":79,"props":304,"children":305},{},[306],{"type":45,"value":307},"The query is clearly extremely simple (fact lookups, single-entity questions)",{"type":39,"tag":79,"props":309,"children":310},{},[311],{"type":45,"value":312},"The query is clearly complex (explicit multi-constraint, \"find everything\", \"exhaustive\", \"comprehensive\")",{"type":39,"tag":79,"props":314,"children":315},{},[316],{"type":45,"value":317},"The user has already specified depth (\"do a deep dive\", \"quick answer\")",{"type":39,"tag":48,"props":319,"children":320},{},[321],{"type":45,"value":322},"Note: if the user explicitly asks for something (e.g. \"100\" of something), continue to work until you've achieved it.",{"type":39,"tag":48,"props":324,"children":325},{},[326,331],{"type":39,"tag":83,"props":327,"children":328},{},[329],{"type":45,"value":330},"What work needs to happen?",{"type":45,"value":332}," Identify which of these apply (most queries use 3-5):",{"type":39,"tag":75,"props":334,"children":335},{},[336,346,356,366,376,386,396,406,416],{"type":39,"tag":79,"props":337,"children":338},{},[339,344],{"type":39,"tag":83,"props":340,"children":341},{},[342],{"type":45,"value":343},"Seed from user input",{"type":45,"value":345},": The user provided a list of entities to start from (company names, tickers, paper titles). Each seed becomes a parallel workstream.",{"type":39,"tag":79,"props":347,"children":348},{},[349,354],{"type":39,"tag":83,"props":350,"children":351},{},[352],{"type":45,"value":353},"Define what qualifies",{"type":45,"value":355},": What makes a result a valid \"row\"? Translate the user's criteria into concrete checks.",{"type":39,"tag":79,"props":357,"children":358},{},[359,364],{"type":39,"tag":83,"props":360,"children":361},{},[362],{"type":45,"value":363},"Define what to capture",{"type":45,"value":365},": What fields (\"columns\") does each result need? Build the schema before searching.",{"type":39,"tag":79,"props":367,"children":368},{},[369,374],{"type":39,"tag":83,"props":370,"children":371},{},[372],{"type":45,"value":373},"Search broadly",{"type":45,"value":375},": Generate diverse queries and run them to find candidates. This is where subagents do the heavy lifting.",{"type":39,"tag":79,"props":377,"children":378},{},[379,384],{"type":39,"tag":83,"props":380,"children":381},{},[382],{"type":45,"value":383},"Extract structured data",{"type":45,"value":385},": Pull specific fields from raw search results into the schema.",{"type":39,"tag":79,"props":387,"children":388},{},[389,394],{"type":39,"tag":83,"props":390,"children":391},{},[392],{"type":45,"value":393},"Filter",{"type":45,"value":395},": Apply hard constraints (dates, geography, thresholds) and soft judgments (quality, relevance, semantic checks).",{"type":39,"tag":79,"props":397,"children":398},{},[399,404],{"type":39,"tag":83,"props":400,"children":401},{},[402],{"type":45,"value":403},"Merge and deduplicate",{"type":45,"value":405},": Combine results from multiple subagents. Same URL = drop duplicate. Same entity from different sources = merge fields, keep best data.",{"type":39,"tag":79,"props":407,"children":408},{},[409,414],{"type":39,"tag":83,"props":410,"children":411},{},[412],{"type":45,"value":413},"Score and rank",{"type":45,"value":415},": For \"best of\" (e.g. \"what's the best ___?\") queries, define the scoring criteria explicitly, then rank.",{"type":39,"tag":79,"props":417,"children":418},{},[419,424],{"type":39,"tag":83,"props":420,"children":421},{},[422],{"type":45,"value":423},"Synthesize narrative",{"type":45,"value":425},": For research queries, organize findings by theme and write prose with citations.",{"type":39,"tag":54,"props":427,"children":429},{"id":428},"step-2-dispatch-subagents",[430],{"type":45,"value":431},"Step 2: Dispatch Subagents",{"type":39,"tag":433,"props":434,"children":436},"h3",{"id":435},"what-subagents-do",[437],{"type":45,"value":438},"What subagents do",{"type":39,"tag":48,"props":440,"children":441},{},[442],{"type":45,"value":443},"Subagents run Exa searches and process the results. They keep raw search output out of your context window. Each subagent should:",{"type":39,"tag":188,"props":445,"children":446},{},[447,452,457],{"type":39,"tag":79,"props":448,"children":449},{},[450],{"type":45,"value":451},"Read the reference file(s) you point it to",{"type":39,"tag":79,"props":453,"children":454},{},[455],{"type":45,"value":456},"Run the specific searches you assign",{"type":39,"tag":79,"props":458,"children":459},{},[460],{"type":45,"value":461},"Return compact, structured output",{"type":39,"tag":433,"props":463,"children":465},{"id":464},"how-to-dispatch",[466],{"type":45,"value":467},"How to dispatch",{"type":39,"tag":48,"props":469,"children":470},{},[471,473,478],{"type":45,"value":472},"Use the ",{"type":39,"tag":83,"props":474,"children":475},{},[476],{"type":45,"value":477},"Agent tool",{"type":45,"value":479}," to dispatch subagents. Reference file paths are relative to the directory this file was loaded from.",{"type":39,"tag":48,"props":481,"children":482},{},[483,485,491],{"type":45,"value":484},"Use ",{"type":39,"tag":66,"props":486,"children":488},{"className":487},[],[489],{"type":45,"value":490},"model: \"haiku\"",{"type":45,"value":492}," for subagents.",{"type":39,"tag":48,"props":494,"children":495},{},[496],{"type":45,"value":497},"Tell each subagent:",{"type":39,"tag":75,"props":499,"children":500},{},[501,506,511],{"type":39,"tag":79,"props":502,"children":503},{},[504],{"type":45,"value":505},"Which reference file(s) to read for instructions (always include the absolute path)",{"type":39,"tag":79,"props":507,"children":508},{},[509],{"type":45,"value":510},"What specific searches to run or what specific work to do",{"type":39,"tag":79,"props":512,"children":513},{},[514],{"type":45,"value":515},"What output format to return",{"type":39,"tag":48,"props":517,"children":518},{},[519],{"type":39,"tag":83,"props":520,"children":521},{},[522],{"type":45,"value":523},"Template:",{"type":39,"tag":525,"props":526,"children":530},"pre",{"className":527,"code":529,"language":45},[528],"language-text","Read the file at [this skill's directory]\u002Freferences\u002Fsearching.md for instructions on how to query Exa effectively.\n\nThen do the following:\n[specific task description]\n[specific queries to run, if you are prescribing them]\n[validation criteria -- what makes a result qualify, so the subagent filters before returning]\n\nReturn: [output format -- e.g. \"compact JSON with name, url, snippet per result\" or \"markdown table with columns X, Y, Z\"].\n\nEnd with EXACTLY: `sources_reviewed: N` where N = sum of `numResults` across every `web_search_exa` call (incl. retries). E.g. calls with numResults 10, 10, 5 → `sources_reviewed: 25`.\n",[531],{"type":39,"tag":66,"props":532,"children":534},{"__ignoreMap":533},"",[535],{"type":45,"value":529},{"type":39,"tag":48,"props":537,"children":538},{},[539],{"type":39,"tag":83,"props":540,"children":541},{},[542,544,550],{"type":45,"value":543},"Pass the ",{"type":39,"tag":66,"props":545,"children":547},{"className":546},[],[548],{"type":45,"value":549},"sources_reviewed",{"type":45,"value":551}," instruction line to every subagent verbatim — don't paraphrase.",{"type":39,"tag":433,"props":553,"children":555},{"id":554},"which-reference-files-to-point-subagents-to",[556],{"type":45,"value":557},"Which reference files to point subagents to",{"type":39,"tag":48,"props":559,"children":560},{},[561,563,568],{"type":45,"value":562},"Always point subagents to ",{"type":39,"tag":66,"props":564,"children":566},{"className":565},[],[567],{"type":45,"value":206},{"type":45,"value":569},". It contains Exa query guidance and an index of domain-specific pattern files that the subagent will select from based on its task.",{"type":39,"tag":48,"props":571,"children":572},{},[573],{"type":45,"value":574},"Point to whichever of these also apply:",{"type":39,"tag":576,"props":577,"children":578},"table",{},[579,598],{"type":39,"tag":580,"props":581,"children":582},"thead",{},[583],{"type":39,"tag":584,"props":585,"children":586},"tr",{},[587,593],{"type":39,"tag":588,"props":589,"children":590},"th",{},[591],{"type":45,"value":592},"File",{"type":39,"tag":588,"props":594,"children":595},{},[596],{"type":45,"value":597},"Point a subagent here when...",{"type":39,"tag":599,"props":600,"children":601},"tbody",{},[602,620,637,654],{"type":39,"tag":584,"props":603,"children":604},{},[605,615],{"type":39,"tag":606,"props":607,"children":608},"td",{},[609],{"type":39,"tag":66,"props":610,"children":612},{"className":611},[],[613],{"type":45,"value":614},"references\u002Fextraction.md",{"type":39,"tag":606,"props":616,"children":617},{},[618],{"type":45,"value":619},"The subagent needs to extract specific data points into a schema you defined",{"type":39,"tag":584,"props":621,"children":622},{},[623,632],{"type":39,"tag":606,"props":624,"children":625},{},[626],{"type":39,"tag":66,"props":627,"children":629},{"className":628},[],[630],{"type":45,"value":631},"references\u002Ffiltering.md",{"type":39,"tag":606,"props":633,"children":634},{},[635],{"type":45,"value":636},"The subagent needs to evaluate results against criteria (especially semantic\u002Fsoft filters)",{"type":39,"tag":584,"props":638,"children":639},{},[640,649],{"type":39,"tag":606,"props":641,"children":642},{},[643],{"type":39,"tag":66,"props":644,"children":646},{"className":645},[],[647],{"type":45,"value":648},"references\u002Fsynthesis.md",{"type":39,"tag":606,"props":650,"children":651},{},[652],{"type":45,"value":653},"The subagent is producing a prose synthesis rather than structured data",{"type":39,"tag":584,"props":655,"children":656},{},[657,666],{"type":39,"tag":606,"props":658,"children":659},{},[660],{"type":39,"tag":66,"props":661,"children":663},{"className":662},[],[664],{"type":45,"value":665},"references\u002Fsource-quality.md",{"type":39,"tag":606,"props":667,"children":668},{},[669],{"type":45,"value":670},"The subagent needs to assess source credibility, especially for \"best of\", ranking, or expert-finding queries",{"type":39,"tag":433,"props":672,"children":674},{"id":673},"how-to-split-work-across-subagents",[675],{"type":45,"value":676},"How to split work across subagents",{"type":39,"tag":48,"props":678,"children":679},{},[680,682,687],{"type":45,"value":681},"If running parallel subagents, decompose the primary task\u002Fquestion into ",{"type":39,"tag":83,"props":683,"children":684},{},[685],{"type":45,"value":686},"sub-questions",{"type":45,"value":688}," to cover different search territories.",{"type":39,"tag":48,"props":690,"children":691},{},[692],{"type":45,"value":693},"For example, \"best open-source LLM fine-tuning frameworks for production use\" can be decomposed into multiple parallel sub-questions:",{"type":39,"tag":75,"props":695,"children":696},{},[697,702,707],{"type":39,"tag":79,"props":698,"children":699},{},[700],{"type":45,"value":701},"\"What open-source LLM fine-tuning frameworks do production engineers recommend, and what do they say about using them in real deployments?\"",{"type":39,"tag":79,"props":703,"children":704},{},[705],{"type":45,"value":706},"\"What open-source LLM fine-tuning tools have launched or gained traction in the last 6 months that aren't yet widely known?\"",{"type":39,"tag":79,"props":708,"children":709},{},[710],{"type":45,"value":711},"\"What are the most common complaints, failure modes, and reasons teams migrated away from specific open-source LLM fine-tuning frameworks in production?\"",{"type":39,"tag":48,"props":713,"children":714},{},[715,717,721],{"type":45,"value":716},"Depending on your \"",{"type":39,"tag":83,"props":718,"children":719},{},[720],{"type":45,"value":186},{"type":45,"value":722},"\" analysis: Some need 2-3; some need many. Some need several different angles, creative thought patterns, adversarial perspectives. It depends on what the user is asking for and how deep they want you to go.",{"type":39,"tag":48,"props":724,"children":725},{},[726],{"type":45,"value":727},"Give the sub-question directly to the subagent in its prompt.",{"type":39,"tag":433,"props":729,"children":731},{"id":730},"subagent-sizing",[732],{"type":45,"value":733},"Subagent sizing",{"type":39,"tag":188,"props":735,"children":736},{},[737,742,747,760],{"type":39,"tag":79,"props":738,"children":739},{},[740],{"type":45,"value":741},"Aim for 3-5 searches per subagent",{"type":39,"tag":79,"props":743,"children":744},{},[745],{"type":45,"value":746},"Parallelize aggressively — independent workstreams should be separate subagents launched in a single message",{"type":39,"tag":79,"props":748,"children":749},{},[750,752,758],{"type":45,"value":751},"Do not use ",{"type":39,"tag":66,"props":753,"children":755},{"className":754},[],[756],{"type":45,"value":757},"run_in_background",{"type":45,"value":759}," — dispatch all subagents in one message and wait for their results",{"type":39,"tag":79,"props":761,"children":762},{},[763],{"type":45,"value":764},"For per-seed work (enriching a list of 20 companies), batch 3-5 seeds per subagent",{"type":39,"tag":433,"props":766,"children":768},{"id":767},"token-isolation",[769],{"type":45,"value":770},"Token isolation",{"type":39,"tag":48,"props":772,"children":773},{},[774],{"type":45,"value":775},"Never run bulk searches in your main context. The whole point of subagents is to keep raw search output out of your context window. Subagents process results and return only distilled output.",{"type":39,"tag":433,"props":777,"children":779},{"id":778},"when-things-go-wrong",[780],{"type":45,"value":781},"When things go wrong",{"type":39,"tag":188,"props":783,"children":784},{},[785,795],{"type":39,"tag":79,"props":786,"children":787},{},[788,793],{"type":39,"tag":83,"props":789,"children":790},{},[791],{"type":45,"value":792},"Subagent returns empty",{"type":45,"value":794},": Rephrase queries with different angles, not synonyms. If still empty, the topic may have limited web coverage -- report that.",{"type":39,"tag":79,"props":796,"children":797},{},[798,803],{"type":39,"tag":83,"props":799,"children":800},{},[801],{"type":45,"value":802},"Subagent returns off-topic results",{"type":45,"value":804},": Queries were too vague. Retry with longer, more specific queries.",{"type":39,"tag":54,"props":806,"children":808},{"id":807},"step-3-compile-results",[809],{"type":45,"value":810},"Step 3: Compile Results",{"type":39,"tag":48,"props":812,"children":813},{},[814],{"type":45,"value":815},"After subagents return:",{"type":39,"tag":48,"props":817,"children":818},{},[819],{"type":39,"tag":83,"props":820,"children":821},{},[822],{"type":45,"value":823},"Deduplicate:",{"type":39,"tag":75,"props":825,"children":826},{},[827,832,837,842],{"type":39,"tag":79,"props":828,"children":829},{},[830],{"type":45,"value":831},"Collect all results into a single list",{"type":39,"tag":79,"props":833,"children":834},{},[835],{"type":45,"value":836},"Remove exact URL duplicates",{"type":39,"tag":79,"props":838,"children":839},{},[840],{"type":45,"value":841},"Same entity from different sources: merge fields, keep the most complete\u002Frecent data",{"type":39,"tag":79,"props":843,"children":844},{},[845],{"type":45,"value":846},"Track: \"Deduplicated X results down to Y unique entries\"",{"type":39,"tag":48,"props":848,"children":849},{},[850],{"type":39,"tag":83,"props":851,"children":852},{},[853],{"type":45,"value":854},"Validate coverage:",{"type":39,"tag":188,"props":856,"children":857},{},[858,863,868],{"type":39,"tag":79,"props":859,"children":860},{},[861],{"type":45,"value":862},"Are there obvious gaps? (missing time periods, missing geographic regions, missing entity types)",{"type":39,"tag":79,"props":864,"children":865},{},[866],{"type":45,"value":867},"For each gap found, run targeted follow-up searches (via subagent if multiple queries are needed, direct if extremely simple)",{"type":39,"tag":79,"props":869,"children":870},{},[871],{"type":45,"value":872},"For \"find everything\" queries, check if results from different subagents overlap heavily (good sign) or are completely disjoint (may indicate missed angles)",{"type":39,"tag":48,"props":874,"children":875},{},[876],{"type":39,"tag":83,"props":877,"children":878},{},[879],{"type":45,"value":880},"Format the output:",{"type":39,"tag":48,"props":882,"children":883},{},[884,886,891],{"type":45,"value":885},"If you used subagents, open with: \"I used Exa to review {X} sources across {Y} subagents. Here's what was found:\" (X = sum of ",{"type":39,"tag":66,"props":887,"children":889},{"className":888},[],[890],{"type":45,"value":549},{"type":45,"value":892}," across all subagents and passes plus any direct searches you ran; Y = total subagents dispatched. Pluralize naturally.)",{"type":39,"tag":48,"props":894,"children":895},{},[896],{"type":45,"value":897},"Then: Format output beautifully, filling up no more than one scroll length of the claude code screen. Include hyperlinked text where relevant. Below it, you may also include things (in a short, easy-to-read format) that:",{"type":39,"tag":188,"props":899,"children":900},{},[901,906,911,916],{"type":39,"tag":79,"props":902,"children":903},{},[904],{"type":45,"value":905},"(\"Result\") directly answer the original user request (in few words; make every word count)",{"type":39,"tag":79,"props":907,"children":908},{},[909],{"type":45,"value":910},"(\"Process\") include anything worth noting about your process and what you consider to be high-signal in this domain vs. what you filtered out.",{"type":39,"tag":79,"props":912,"children":913},{},[914],{"type":45,"value":915},"(\"Patterns\") any patterns identified that are non-obvious, require n-th order thinking, and are not included or alluded to in the rest of the output but might be interesting to the user.",{"type":39,"tag":79,"props":917,"children":918},{},[919],{"type":45,"value":920},"(\"Notes\") based on everything you know about the user and their work beyond this task, mention anything notable\u002Fuseful you found that is not included or alluded to in the rest of the output.",{"type":39,"tag":48,"props":922,"children":923},{},[924,926,932],{"type":45,"value":925},"If it's impossible to fit the full output in a single screen, write a file in the most relevant\u002Fuseful file format (.csv, .md) to ",{"type":39,"tag":66,"props":927,"children":929},{"className":928},[],[930],{"type":45,"value":931},".\u002Fexa-results\u002F\u003Ctopic>-\u003CYYYY-MM-DD>",{"type":45,"value":933}," and include a pointer to the full file below the 1-screen output.",{"type":39,"tag":48,"props":935,"children":936},{},[937],{"type":39,"tag":83,"props":938,"children":939},{},[940],{"type":45,"value":941},"General output rules:",{"type":39,"tag":188,"props":943,"children":944},{},[945,950,955],{"type":39,"tag":79,"props":946,"children":947},{},[948],{"type":45,"value":949},"No emojis unless the user requested them",{"type":39,"tag":79,"props":951,"children":952},{},[953],{"type":45,"value":954},"Include in-line 1-word or multi-word hyperlinks throughout outputs where hyperlinking is a value-add.",{"type":39,"tag":79,"props":956,"children":957},{},[958],{"type":45,"value":959},"Prefer tables over lists (fall back to lists only when fields are non-uniform or values are too long to fit cleanly)",{"type":39,"tag":54,"props":961,"children":963},{"id":962},"multi-pass-queries",[964],{"type":45,"value":965},"Multi-Pass Queries",{"type":39,"tag":48,"props":967,"children":968},{},[969],{"type":45,"value":970},"Some queries require multiple sequential passes where later passes depend on earlier results. Common patterns:",{"type":39,"tag":48,"props":972,"children":973},{},[974,979],{"type":39,"tag":83,"props":975,"children":976},{},[977],{"type":45,"value":978},"Entity chaining",{"type":45,"value":980}," (multi-hop): Pass 1 finds entities (companies), Pass 2 finds related entities per result (people at those companies), Pass 3 enriches those (their public statements). Each pass is a round of parallel subagents.",{"type":39,"tag":48,"props":982,"children":983},{},[984,989],{"type":39,"tag":83,"props":985,"children":986},{},[987],{"type":45,"value":988},"Exploratory then targeted",{"type":45,"value":990},": Pass 1 scouts the landscape broadly, Pass 2 searches deeply in the most promising directions found in Pass 1.",{"type":39,"tag":48,"props":992,"children":993},{},[994,999],{"type":39,"tag":83,"props":995,"children":996},{},[997],{"type":45,"value":998},"Criteria discovery",{"type":45,"value":1000},": When \"best\" isn't predefined, Pass 1 surveys what practitioners actually value, Pass 2 searches for candidates matching those criteria.",{"type":39,"tag":48,"props":1002,"children":1003},{},[1004],{"type":45,"value":1005},"Between passes, compile and deduplicate before dispatching the next round.",{"type":39,"tag":54,"props":1007,"children":1009},{"id":1008},"evaluating-source-quality",[1010],{"type":45,"value":1011},"Evaluating Source Quality",{"type":39,"tag":48,"props":1013,"children":1014},{},[1015],{"type":45,"value":1016},"Source quality matters most for \"best of\", ranking, expert-finding, and best-practices queries, but is useful context for almost any research task.",{"type":39,"tag":48,"props":1018,"children":1019},{},[1020,1025,1027,1032],{"type":39,"tag":83,"props":1021,"children":1022},{},[1023],{"type":45,"value":1024},"At the subagent level:",{"type":45,"value":1026}," Point subagents to ",{"type":39,"tag":66,"props":1028,"children":1030},{"className":1029},[],[1031],{"type":45,"value":665},{"type":45,"value":1033}," so they tag source quality in their output. This lets you weight results during compilation.",{"type":39,"tag":48,"props":1035,"children":1036},{},[1037,1042],{"type":39,"tag":83,"props":1038,"children":1039},{},[1040],{"type":45,"value":1041},"At the orchestrator level",{"type":45,"value":1043},", when compiling subagent results:",{"type":39,"tag":75,"props":1045,"children":1046},{},[1047,1057,1067,1077,1087],{"type":39,"tag":79,"props":1048,"children":1049},{},[1050,1055],{"type":39,"tag":83,"props":1051,"children":1052},{},[1053],{"type":45,"value":1054},"Convergence across high-signal sources",{"type":45,"value":1056},": Convergence alone isn't meaningful (3 low-quality sources agreeing is just shared noise). What matters is when multiple independent, high-signal sources (practitioners, people with skin in the game) converge on the same finding.",{"type":39,"tag":79,"props":1058,"children":1059},{},[1060,1065],{"type":39,"tag":83,"props":1061,"children":1062},{},[1063],{"type":45,"value":1064},"Practitioner vs commentator",{"type":45,"value":1066},": Weight practitioners (people doing the work) higher than commentators (people writing about the work).",{"type":39,"tag":79,"props":1068,"children":1069},{},[1070,1075],{"type":39,"tag":83,"props":1071,"children":1072},{},[1073],{"type":45,"value":1074},"Via negativa",{"type":45,"value":1076},": Before synthesizing, define who to exclude (sources with misaligned incentives, no skin in the game, or unfalsifiable claims). Filtering out noise is more valuable than seeking brilliance.",{"type":39,"tag":79,"props":1078,"children":1079},{},[1080,1085],{"type":39,"tag":83,"props":1081,"children":1082},{},[1083],{"type":45,"value":1084},"Red-team your compiled results",{"type":45,"value":1086},": What perspectives are missing? What biases might be distorting the aggregate? If a gap emerges, run a targeted follow-up.",{"type":39,"tag":79,"props":1088,"children":1089},{},[1090,1095],{"type":39,"tag":83,"props":1091,"children":1092},{},[1093],{"type":45,"value":1094},"Ideas over entities",{"type":45,"value":1096},": For expert-finding and best-practices queries, the primary output is convergent truths, not a ranked list of names. Lead with what the best sources agree on, then cite who said it.",{"type":39,"tag":54,"props":1098,"children":1100},{"id":1099},"gotchas",[1101],{"type":45,"value":1102},"Gotchas",{"type":39,"tag":188,"props":1104,"children":1105},{},[1106,1116,1126,1136,1146,1156],{"type":39,"tag":79,"props":1107,"children":1108},{},[1109,1114],{"type":39,"tag":83,"props":1110,"children":1111},{},[1112],{"type":45,"value":1113},"Over-execution on simple queries",{"type":45,"value":1115},": If the user asks \"what year was X founded\", don't spin up subagents. One search, one answer.",{"type":39,"tag":79,"props":1117,"children":1118},{},[1119,1124],{"type":39,"tag":83,"props":1120,"children":1121},{},[1122],{"type":45,"value":1123},"Under-execution on hard queries",{"type":45,"value":1125},": If the query has 4+ constraints, temporal joins, or semantic filtering, a single search will not cut it. Fan out.",{"type":39,"tag":79,"props":1127,"children":1128},{},[1129,1134],{"type":39,"tag":83,"props":1130,"children":1131},{},[1132],{"type":45,"value":1133},"Synonym queries",{"type":45,"value":1135},": Running \"overrated AI tools\" and \"overhyped AI tools\" as separate subagent queries wastes tokens. These hit the same embedding region. Diversify by angle instead.",{"type":39,"tag":79,"props":1137,"children":1138},{},[1139,1144],{"type":39,"tag":83,"props":1140,"children":1141},{},[1142],{"type":45,"value":1143},"Forgetting to deduplicate",{"type":45,"value":1145},": Multiple subagents will return overlapping results. Always deduplicate before synthesis.",{"type":39,"tag":79,"props":1147,"children":1148},{},[1149,1154],{"type":39,"tag":83,"props":1150,"children":1151},{},[1152],{"type":45,"value":1153},"Treating Exa results as validated",{"type":45,"value":1155},": Exa returns similarity, not yet validated. A result appearing in search output does not mean it meets the user's criteria. You must validate.",{"type":39,"tag":79,"props":1157,"children":1158},{},[1159,1164],{"type":39,"tag":83,"props":1160,"children":1161},{},[1162],{"type":45,"value":1163},"Date drift",{"type":45,"value":1165},": Always calculate dates from the current environment date. Never reuse dates from these instructions or from previous queries.",{"items":1167,"total":1290},[1168,1187,1207,1226,1243,1259,1278],{"slug":1169,"name":1169,"fn":1170,"description":1171,"org":1172,"tags":1173,"stars":22,"repoUrl":23,"updatedAt":1186},"amazon-location-service","integrate Amazon Location Service APIs","Integrates Amazon Location Service APIs for AWS applications. Use this skill when users want to add maps (interactive MapLibre or static images); geocode addresses to coordinates or reverse geocode coordinates to addresses; calculate routes, travel times, or service areas; find places and businesses through text search, nearby search, or autocomplete suggestions; retrieve detailed place information including hours, contacts, and addresses; monitor geographical boundaries with geofences; or track device locations. Covers authentication, SDK integration, and all Amazon Location Service capabilities.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1174,1177,1180,1183],{"name":1175,"slug":1176,"type":15},"API Development","api-development",{"name":1178,"slug":1179,"type":15},"AWS","aws",{"name":1181,"slug":1182,"type":15},"Maps","maps",{"name":1184,"slug":1185,"type":15},"Navigation","navigation","2026-07-12T08:13:53.294026",{"slug":1188,"name":1188,"fn":1189,"description":1190,"org":1191,"tags":1192,"stars":22,"repoUrl":23,"updatedAt":1206},"amplify-workflow","build full-stack apps with AWS Amplify","Build and deploy full-stack web and mobile apps with AWS Amplify Gen2 (TypeScript code-first). Covers auth (Cognito), data (AppSync\u002FDynamoDB including schema modeling, enum types, relationships, authorization rules), storage (S3), functions, APIs, and AI (Amplify AI Kit with Bedrock). Supports React, Next.js, Vue, Angular, React Native, Flutter, Swift, and Android. Always use this skill for Amplify Gen2 topics — even for questions you think you know — it contains validated, version-specific patterns that prevent common mistakes. TRIGGER when: user mentions Amplify Gen2; project has amplify\u002F directory or amplify_outputs; code imports @aws-amplify packages; user asks about defineBackend, defineAuth, defineData, defineStorage, or npx ampx. SKIP: Amplify Gen1 (amplify CLI v6), standalone SAM\u002FCDK without Amplify (use aws-serverless), direct Bedrock without Amplify AI Kit (use bedrock).",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1193,1196,1197,1200,1203],{"name":1194,"slug":1195,"type":15},"Auth","auth",{"name":1178,"slug":1179,"type":15},{"name":1198,"slug":1199,"type":15},"Database","database",{"name":1201,"slug":1202,"type":15},"Frontend","frontend",{"name":1204,"slug":1205,"type":15},"TypeScript","typescript","2026-07-12T08:13:47.134012",{"slug":1208,"name":1208,"fn":1209,"description":1210,"org":1211,"tags":1212,"stars":22,"repoUrl":23,"updatedAt":1225},"analyzer","analyze queried data for trends","Analyze queried data for trends, week-over-week comparisons, distributions, funnels, cohorts, top-N lists, anomalies, sanity checks, and report-ready findings. Use after or alongside ClickHouse queries when the user wants insight rather than raw rows.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1213,1216,1219,1222],{"name":1214,"slug":1215,"type":15},"Analytics","analytics",{"name":1217,"slug":1218,"type":15},"ClickHouse","clickhouse",{"name":1220,"slug":1221,"type":15},"Data Analysis","data-analysis",{"name":1223,"slug":1224,"type":15},"Statistics","statistics","2026-07-12T08:14:05.606036",{"slug":1227,"name":1227,"fn":1228,"description":1229,"org":1230,"tags":1231,"stars":22,"repoUrl":23,"updatedAt":1242},"artifact-management","manage and organize analysis artifacts","Save, organize, and describe reusable analysis artifacts such as SQL, result snapshots, CSV exports, summaries, caveats, plots, and report-ready files. Use when users ask to save, export, share, cite, reproduce, or organize data-analysis outputs.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1232,1233,1236,1239],{"name":1220,"slug":1221,"type":15},{"name":1234,"slug":1235,"type":15},"Productivity","productivity",{"name":1237,"slug":1238,"type":15},"Reporting","reporting",{"name":1240,"slug":1241,"type":15},"SQL","sql","2026-07-12T08:14:09.265555",{"slug":1244,"name":1244,"fn":1245,"description":1246,"org":1247,"tags":1248,"stars":22,"repoUrl":23,"updatedAt":1258},"attorney-assist","connect with attorneys for legal consultation","Connects the user with a LegalZoom attorney for legal consultation. Use when a user asks about attorneys, lawyers, or legal help, or when contract review reveals high risks or low-confidence findings.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1249,1252,1255],{"name":1250,"slug":1251,"type":15},"Contracts","contracts",{"name":1253,"slug":1254,"type":15},"Legal","legal",{"name":1256,"slug":1257,"type":15},"Risk Assessment","risk-assessment","2026-07-12T08:13:45.878361",{"slug":1260,"name":1260,"fn":1261,"description":1262,"org":1263,"tags":1264,"stars":22,"repoUrl":23,"updatedAt":1277},"building-pydantic-ai-agents","build AI agents with Pydantic AI","Build AI agents with Pydantic AI — tools, capabilities (including on-demand loading), structured output, streaming, testing, and multi-agent patterns. Use when the user mentions Pydantic AI, imports pydantic_ai, or asks to build an AI agent, add tools\u002Fcapabilities, defer capability loading, stream output, define agents from YAML, or test agent behavior.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1265,1268,1271,1274],{"name":1266,"slug":1267,"type":15},"Agents","agents",{"name":1269,"slug":1270,"type":15},"LLM","llm",{"name":1272,"slug":1273,"type":15},"Multi-Agent","multi-agent",{"name":1275,"slug":1276,"type":15},"Python","python","2026-07-12T08:14:01.893781",{"slug":1218,"name":1218,"fn":1279,"description":1280,"org":1281,"tags":1282,"stars":22,"repoUrl":23,"updatedAt":1289},"query ClickHouse databases via CLI","Connect to and query ClickHouse (a local server or a ClickHouse Cloud service) from the terminal using the official clickhousectl CLI, including the browser OAuth login flow. Use when the user wants to run SQL against ClickHouse, explore schemas and tables, inspect Cloud services, or authenticate clickhousectl. For building a local dev environment or deploying to Cloud, defer to the official ClickHouse skills (see Scope).",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1283,1284,1287,1288],{"name":1214,"slug":1215,"type":15},{"name":1285,"slug":1286,"type":15},"CLI","cli",{"name":1217,"slug":1218,"type":15},{"name":1198,"slug":1199,"type":15},"2026-07-12T08:14:06.829692",43,{"items":1292,"total":1418},[1293,1300,1308,1315,1322,1328,1335,1342,1354,1372,1392,1405],{"slug":1169,"name":1169,"fn":1170,"description":1171,"org":1294,"tags":1295,"stars":22,"repoUrl":23,"updatedAt":1186},{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1296,1297,1298,1299],{"name":1175,"slug":1176,"type":15},{"name":1178,"slug":1179,"type":15},{"name":1181,"slug":1182,"type":15},{"name":1184,"slug":1185,"type":15},{"slug":1188,"name":1188,"fn":1189,"description":1190,"org":1301,"tags":1302,"stars":22,"repoUrl":23,"updatedAt":1206},{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1303,1304,1305,1306,1307],{"name":1194,"slug":1195,"type":15},{"name":1178,"slug":1179,"type":15},{"name":1198,"slug":1199,"type":15},{"name":1201,"slug":1202,"type":15},{"name":1204,"slug":1205,"type":15},{"slug":1208,"name":1208,"fn":1209,"description":1210,"org":1309,"tags":1310,"stars":22,"repoUrl":23,"updatedAt":1225},{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1311,1312,1313,1314],{"name":1214,"slug":1215,"type":15},{"name":1217,"slug":1218,"type":15},{"name":1220,"slug":1221,"type":15},{"name":1223,"slug":1224,"type":15},{"slug":1227,"name":1227,"fn":1228,"description":1229,"org":1316,"tags":1317,"stars":22,"repoUrl":23,"updatedAt":1242},{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1318,1319,1320,1321],{"name":1220,"slug":1221,"type":15},{"name":1234,"slug":1235,"type":15},{"name":1237,"slug":1238,"type":15},{"name":1240,"slug":1241,"type":15},{"slug":1244,"name":1244,"fn":1245,"description":1246,"org":1323,"tags":1324,"stars":22,"repoUrl":23,"updatedAt":1258},{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1325,1326,1327],{"name":1250,"slug":1251,"type":15},{"name":1253,"slug":1254,"type":15},{"name":1256,"slug":1257,"type":15},{"slug":1260,"name":1260,"fn":1261,"description":1262,"org":1329,"tags":1330,"stars":22,"repoUrl":23,"updatedAt":1277},{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1331,1332,1333,1334],{"name":1266,"slug":1267,"type":15},{"name":1269,"slug":1270,"type":15},{"name":1272,"slug":1273,"type":15},{"name":1275,"slug":1276,"type":15},{"slug":1218,"name":1218,"fn":1279,"description":1280,"org":1336,"tags":1337,"stars":22,"repoUrl":23,"updatedAt":1289},{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1338,1339,1340,1341],{"name":1214,"slug":1215,"type":15},{"name":1285,"slug":1286,"type":15},{"name":1217,"slug":1218,"type":15},{"name":1198,"slug":1199,"type":15},{"slug":1343,"name":1343,"fn":1344,"description":1345,"org":1346,"tags":1347,"stars":22,"repoUrl":23,"updatedAt":1353},"cline-session-history","search and browse Cline session history","Search and browse Cline session history. Use when the user asks to find, list, inspect, or export Cline sessions by time period, prompt content, status, model, provider, source, mode, workspace, or other criteria. Also use when the user wants to read a session transcript or export sessions to a directory. Trigger phrases include \"find my session\", \"search my session history\", \"show me past sessions\", \"what was that session where\", \"find the session that started with\", and any mention of past Cline conversations.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1348,1349,1352],{"name":1266,"slug":1267,"type":15},{"name":1350,"slug":1351,"type":15},"History","history",{"name":1234,"slug":1235,"type":15},"2026-07-19T06:03:13.945151",{"slug":1355,"name":1355,"fn":1356,"description":1357,"org":1358,"tags":1359,"stars":22,"repoUrl":23,"updatedAt":1371},"convex-design","build reactive backends with Convex","Design and build reactive, type-safe, production-grade backends on Convex. Covers schema, queries\u002Fmutations\u002Factions, indexes, auth, file storage, scheduling, real-time multiplayer, mobile backends, and LLM\u002Fagent workflows on Convex's one-platform stack.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1360,1361,1364,1365,1368],{"name":1194,"slug":1195,"type":15},{"name":1362,"slug":1363,"type":15},"Backend","backend",{"name":1198,"slug":1199,"type":15},{"name":1366,"slug":1367,"type":15},"Real-time","real-time",{"name":1369,"slug":1370,"type":15},"Storage","storage","2026-07-12T08:13:37.101253",{"slug":1373,"name":1373,"fn":1374,"description":1375,"org":1376,"tags":1377,"stars":22,"repoUrl":23,"updatedAt":1391},"cosmosdb-best-practices","optimize Azure Cosmos DB performance","Azure Cosmos DB performance optimization and best practices guidelines for NoSQL,\npartitioning, queries, SDK usage, and vector search. Use when writing, reviewing,\nor refactoring code that interacts with Azure Cosmos DB, designing data models,\noptimizing queries, or implementing high-performance database operations.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1378,1381,1384,1385,1388],{"name":1379,"slug":1380,"type":15},"Azure","azure",{"name":1382,"slug":1383,"type":15},"Cosmos DB","cosmos-db",{"name":1198,"slug":1199,"type":15},{"name":1386,"slug":1387,"type":15},"NoSQL","nosql",{"name":1389,"slug":1390,"type":15},"Performance","performance","2026-07-12T08:13:54.531719",{"slug":1393,"name":1393,"fn":1394,"description":1395,"org":1396,"tags":1397,"stars":22,"repoUrl":23,"updatedAt":1404},"data-analyst","analyze ClickHouse analytics data","Act as an interactive data analyst for ClickHouse-backed analytics. Use when the user asks questions about internal data, metrics, dashboards, telemetry, active users, revenue, funnels, trends, distributions, or wants an analyst-style conversation, ad hoc SQL, charts, or a data export against ClickHouse (local or ClickHouse Cloud).",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1398,1399,1400,1403],{"name":1214,"slug":1215,"type":15},{"name":1217,"slug":1218,"type":15},{"name":1401,"slug":1402,"type":15},"Dashboards","dashboards",{"name":1220,"slug":1221,"type":15},"2026-07-12T08:13:31.975246",{"slug":1406,"name":1406,"fn":1407,"description":1408,"org":1409,"tags":1410,"stars":22,"repoUrl":23,"updatedAt":1417},"dataproc-skills","manage Dataproc clusters and jobs","Skills to interact with your Dataproc clusters and jobs.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1411,1414],{"name":1412,"slug":1413,"type":15},"Data Engineering","data-engineering",{"name":1415,"slug":1416,"type":15},"Operations","operations","2026-07-12T08:13:42.179275",45]