[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-posthog-improving-mcp-tools":3,"mdc--msuoq2-key":52,"related-repo-posthog-improving-mcp-tools":496,"related-org-posthog-improving-mcp-tools":594},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":26,"repoUrl":27,"updatedAt":28,"license":29,"forks":30,"topics":31,"repo":47,"sourceUrl":50,"mdContent":51},"improving-mcp-tools","improve MCP tool performance via evaluation","Run an improve-my-MCP campaign: an autoresearch-style loop that measures the MCP agent experience with the eval harness, picks the highest-impact tool problem from production data, makes one bounded fix, and keeps it only if before\u002Fafter scores improve. Use when asked to \"improve my MCP\", run an MCP improvement campaign, fix tool discoverability or descriptions based on evidence, or prepare an eval-backed PR for a tool change. Every shipped change must carry eval evidence; guardrails below are hard rules.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},"posthog","PostHog","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fposthog.png",[12,16,17,20,23],{"name":13,"slug":14,"type":15},"Performance","performance","tag",{"name":9,"slug":8,"type":15},{"name":18,"slug":19,"type":15},"Evals","evals",{"name":21,"slug":22,"type":15},"MCP","mcp",{"name":24,"slug":25,"type":15},"Debugging","debugging",35568,"https:\u002F\u002Fgithub.com\u002FPostHog\u002Fposthog","2026-07-04T07:09:26.072517",null,2977,[32,33,34,35,36,37,38,39,40,41,42,43,44,45,46],"ab-testing","ai-analytics","analytics","cdp","data-warehouse","experiments","feature-flags","javascript","product-analytics","python","react","session-replay","surveys","typescript","web-analytics",{"repoUrl":27,"stars":26,"forks":30,"topics":48,"description":49},[32,33,34,35,36,37,38,39,40,41,42,43,44,45,46],"🦔 PostHog is an all-in-one developer platform for building successful products. We offer product analytics, web analytics, session replay, error tracking, feature flags, experimentation, surveys, data warehouse, a CDP, and an AI product assistant to help debug your code, ship features faster, and keep all your usage and customer data in one stack.","https:\u002F\u002Fgithub.com\u002FPostHog\u002Fposthog\u002Ftree\u002FHEAD\u002Fproducts\u002Fmcp_analytics\u002Fskills\u002Fimproving-mcp-tools","---\nname: improving-mcp-tools\ndescription: >\n  Run an improve-my-MCP campaign: an autoresearch-style loop that measures the\n  MCP agent experience with the eval harness, picks the highest-impact tool\n  problem from production data, makes one bounded fix, and keeps it only if\n  before\u002Fafter scores improve. Use when asked to \"improve my MCP\", run an MCP\n  improvement campaign, fix tool discoverability or descriptions based on\n  evidence, or prepare an eval-backed PR for a tool change. Every shipped\n  change must carry eval evidence; guardrails below are hard rules.\n---\n\n# Improving MCP tools\n\nAn MCP server gets better only in ways you can measure. This skill is the\ncampaign procedure: score the current agent experience, fix the biggest\nproblem, re-score, and only ship changes the numbers justify. It is the\noperating manual for the \"improve my MCP\" loop — one iteration per pass,\njournaled so a later iteration (or a different agent) can resume without\nrepeating work.\n\n## The objective function\n\n`services\u002Fmcp\u002Fevals\u002F` is the harness. `benchmark\u002Ftasks.yaml` is a fixed set of\nagent tasks with `expected_tools` and `success_criteria`; scores are only\ncomparable across runs of the same benchmark `version`.\n\n- **Probe mode** (deterministic, no LLM):\n  `LIVE_MCP_URL=... LIVE_MCP_TOKEN=... pnpm exec tsx evals\u002Frunner\u002Fprobe.ts --out score.json`\n  from `services\u002Fmcp\u002F`. Reports tool-presence misses (discoverability), probe\n  failures, and latency p50\u002Fp95. Non-zero exit = regression.\n- **Agent mode** (LLM replay + judge): scores task success and tool-selection\n  accuracy. Use it for description\u002Fdiscoverability changes — probes cannot\n  detect that an agent picks the wrong tool.\n\nRun the harness against a **seeded local or devbox stack**, never against a\ncustomer project. Local recipe: `NODE_ENV=development PORT=9876\nPOSTHOG_API_BASE_URL=http:\u002F\u002Flocalhost:8000 pnpm dev:hono`, personal API key as\n`LIVE_MCP_TOKEN`.\n\n## One iteration\n\n1. **Measure.** Run the harness for a baseline. Pull production evidence with\n   the MCP analytics tools (`query-mcp-tool-stats`, `query-mcp-tool-failures`,\n   `query-mcp-tool-descriptions`, `query-mcp-tool-sample-intents`) and the\n   lenses in the signals scout cookbook\n   (`products\u002Fsignals\u002Fskills\u002Fsignals-scout-mcp-tool-calls\u002Freferences\u002Fqueries.md`):\n   failure leaderboard, retry\u002Fstruggle, latency, intents that matched no tool.\n2. **Pick one issue.** Rank by reach × severity. Skip anything the journal\n   shows with two failed attempts. One issue per iteration — a PR that fixes\n   three things can't be attributed to any of them when scores move.\n3. **Fix, bounded.** Only files inside the allowlist (below). Typical fixes:\n   sharpen a tool description so the right intent finds it, tighten an input\n   schema that agents keep getting wrong, fix an annotation, update a skill.\n4. **Validate.** Re-run the affected benchmark slice plus a no-regression\n   sample. Keep the change only if the target metric improves and nothing else\n   degrades. A discarded change is a normal outcome — journal it and move on.\n5. **Ship.** One PR per iteration with before\u002Fafter scores in the body (format\n   in [references\u002Fcampaign-journal.md](references\u002Fcampaign-journal.md)). Keep\n   it stampable: ≤400 changed lines, only files inside the allowlist below,\n   apply the `stamphog` label. Autonomy level comes from the campaign config —\n   default is **draft PR for human review**; only arm auto-merge when the\n   operator has explicitly enabled the self-driving experiment (see\n   guardrails).\n6. **Journal.** Append the iteration record before ending the pass.\n\n## Hard guardrails\n\nThese are not suggestions; violating any of them ends the campaign pass.\n\n- **Allowlist** — a campaign PR may only touch: `products\u002F*\u002Fmcp\u002Ftools.yaml`,\n  `products\u002F*\u002Fskills\u002F**`, `services\u002Fmcp\u002Fevals\u002F**`, the codegen outputs of\n  `pnpm generate-tools` \u002F `scaffold-yaml` (`services\u002Fmcp\u002Fsrc\u002Ftools\u002Fgenerated\u002F**`\n  and `services\u002Fmcp\u002Fschema\u002Fgenerated-tool-definitions.json`), and docs.\n  Anything else (handler code, package manifests, workflows, migrations, auth\n  paths) → stop and hand the finding to a human as a draft PR or report\n  instead.\n- **Read-only against data.** The harness and all production queries are\n  read-only. Never create, mutate, or delete customer-visible objects while\n  measuring.\n- **Evidence or it didn't happen.** No PR without a baseline score, an after\n  score, and the exact harness commands used.\n- **Benchmark integrity.** Never edit `benchmark\u002Ftasks.yaml` in the same PR as\n  a fix it validates — changing the exam and the answer together proves\n  nothing. Benchmark changes are their own PR and bump `version`.\n- **Budgets.** Respect the operator's iteration\u002Ftoken\u002FPR caps (default: stop\n  after 3 open unmerged campaign PRs). Two failed attempts on an issue parks\n  it permanently.\n- **Kill switch.** If the campaign config, its feature flag, or the operator\n  says stop — stop mid-iteration, journal state, end cleanly.\n\n## Failure modes to expect\n\n- A description change that helps one intent can steal traffic from the right\n  tool for another — that's why the no-regression sample is mandatory.\n- Probe latency varies with stack warmth; compare medians across ≥3 runs\n  before attributing a latency change to your fix.\n- Tool-presence misses can be feature-flag gating, not catalog absence —\n  check `getToolsForFeatures` gating before \"fixing\" discoverability.\n",{"data":53,"body":54},{"name":4,"description":6},{"type":55,"children":56},"root",[57,65,71,78,122,164,191,197,323,329,334,464,470],{"type":58,"tag":59,"props":60,"children":61},"element","h1",{"id":4},[62],{"type":63,"value":64},"text","Improving MCP tools",{"type":58,"tag":66,"props":67,"children":68},"p",{},[69],{"type":63,"value":70},"An MCP server gets better only in ways you can measure. This skill is the\ncampaign procedure: score the current agent experience, fix the biggest\nproblem, re-score, and only ship changes the numbers justify. It is the\noperating manual for the \"improve my MCP\" loop — one iteration per pass,\njournaled so a later iteration (or a different agent) can resume without\nrepeating work.",{"type":58,"tag":72,"props":73,"children":75},"h2",{"id":74},"the-objective-function",[76],{"type":63,"value":77},"The objective function",{"type":58,"tag":66,"props":79,"children":80},{},[81,88,90,96,98,104,106,112,114,120],{"type":58,"tag":82,"props":83,"children":85},"code",{"className":84},[],[86],{"type":63,"value":87},"services\u002Fmcp\u002Fevals\u002F",{"type":63,"value":89}," is the harness. ",{"type":58,"tag":82,"props":91,"children":93},{"className":92},[],[94],{"type":63,"value":95},"benchmark\u002Ftasks.yaml",{"type":63,"value":97}," is a fixed set of\nagent tasks with ",{"type":58,"tag":82,"props":99,"children":101},{"className":100},[],[102],{"type":63,"value":103},"expected_tools",{"type":63,"value":105}," and ",{"type":58,"tag":82,"props":107,"children":109},{"className":108},[],[110],{"type":63,"value":111},"success_criteria",{"type":63,"value":113},"; scores are only\ncomparable across runs of the same benchmark ",{"type":58,"tag":82,"props":115,"children":117},{"className":116},[],[118],{"type":63,"value":119},"version",{"type":63,"value":121},".",{"type":58,"tag":123,"props":124,"children":125},"ul",{},[126,154],{"type":58,"tag":127,"props":128,"children":129},"li",{},[130,136,138,144,146,152],{"type":58,"tag":131,"props":132,"children":133},"strong",{},[134],{"type":63,"value":135},"Probe mode",{"type":63,"value":137}," (deterministic, no LLM):\n",{"type":58,"tag":82,"props":139,"children":141},{"className":140},[],[142],{"type":63,"value":143},"LIVE_MCP_URL=... LIVE_MCP_TOKEN=... pnpm exec tsx evals\u002Frunner\u002Fprobe.ts --out score.json",{"type":63,"value":145},"\nfrom ",{"type":58,"tag":82,"props":147,"children":149},{"className":148},[],[150],{"type":63,"value":151},"services\u002Fmcp\u002F",{"type":63,"value":153},". Reports tool-presence misses (discoverability), probe\nfailures, and latency p50\u002Fp95. Non-zero exit = regression.",{"type":58,"tag":127,"props":155,"children":156},{},[157,162],{"type":58,"tag":131,"props":158,"children":159},{},[160],{"type":63,"value":161},"Agent mode",{"type":63,"value":163}," (LLM replay + judge): scores task success and tool-selection\naccuracy. Use it for description\u002Fdiscoverability changes — probes cannot\ndetect that an agent picks the wrong tool.",{"type":58,"tag":66,"props":165,"children":166},{},[167,169,174,176,182,184,190],{"type":63,"value":168},"Run the harness against a ",{"type":58,"tag":131,"props":170,"children":171},{},[172],{"type":63,"value":173},"seeded local or devbox stack",{"type":63,"value":175},", never against a\ncustomer project. Local recipe: ",{"type":58,"tag":82,"props":177,"children":179},{"className":178},[],[180],{"type":63,"value":181},"NODE_ENV=development PORT=9876 POSTHOG_API_BASE_URL=http:\u002F\u002Flocalhost:8000 pnpm dev:hono",{"type":63,"value":183},", personal API key as\n",{"type":58,"tag":82,"props":185,"children":187},{"className":186},[],[188],{"type":63,"value":189},"LIVE_MCP_TOKEN",{"type":63,"value":121},{"type":58,"tag":72,"props":192,"children":194},{"id":193},"one-iteration",[195],{"type":63,"value":196},"One iteration",{"type":58,"tag":198,"props":199,"children":200},"ol",{},[201,250,260,270,280,313],{"type":58,"tag":127,"props":202,"children":203},{},[204,209,211,217,219,225,227,233,234,240,242,248],{"type":58,"tag":131,"props":205,"children":206},{},[207],{"type":63,"value":208},"Measure.",{"type":63,"value":210}," Run the harness for a baseline. Pull production evidence with\nthe MCP analytics tools (",{"type":58,"tag":82,"props":212,"children":214},{"className":213},[],[215],{"type":63,"value":216},"query-mcp-tool-stats",{"type":63,"value":218},", ",{"type":58,"tag":82,"props":220,"children":222},{"className":221},[],[223],{"type":63,"value":224},"query-mcp-tool-failures",{"type":63,"value":226},",\n",{"type":58,"tag":82,"props":228,"children":230},{"className":229},[],[231],{"type":63,"value":232},"query-mcp-tool-descriptions",{"type":63,"value":218},{"type":58,"tag":82,"props":235,"children":237},{"className":236},[],[238],{"type":63,"value":239},"query-mcp-tool-sample-intents",{"type":63,"value":241},") and the\nlenses in the signals scout cookbook\n(",{"type":58,"tag":82,"props":243,"children":245},{"className":244},[],[246],{"type":63,"value":247},"products\u002Fsignals\u002Fskills\u002Fsignals-scout-mcp-tool-calls\u002Freferences\u002Fqueries.md",{"type":63,"value":249},"):\nfailure leaderboard, retry\u002Fstruggle, latency, intents that matched no tool.",{"type":58,"tag":127,"props":251,"children":252},{},[253,258],{"type":58,"tag":131,"props":254,"children":255},{},[256],{"type":63,"value":257},"Pick one issue.",{"type":63,"value":259}," Rank by reach × severity. Skip anything the journal\nshows with two failed attempts. One issue per iteration — a PR that fixes\nthree things can't be attributed to any of them when scores move.",{"type":58,"tag":127,"props":261,"children":262},{},[263,268],{"type":58,"tag":131,"props":264,"children":265},{},[266],{"type":63,"value":267},"Fix, bounded.",{"type":63,"value":269}," Only files inside the allowlist (below). Typical fixes:\nsharpen a tool description so the right intent finds it, tighten an input\nschema that agents keep getting wrong, fix an annotation, update a skill.",{"type":58,"tag":127,"props":271,"children":272},{},[273,278],{"type":58,"tag":131,"props":274,"children":275},{},[276],{"type":63,"value":277},"Validate.",{"type":63,"value":279}," Re-run the affected benchmark slice plus a no-regression\nsample. Keep the change only if the target metric improves and nothing else\ndegrades. A discarded change is a normal outcome — journal it and move on.",{"type":58,"tag":127,"props":281,"children":282},{},[283,288,290,296,298,304,306,311],{"type":58,"tag":131,"props":284,"children":285},{},[286],{"type":63,"value":287},"Ship.",{"type":63,"value":289}," One PR per iteration with before\u002Fafter scores in the body (format\nin ",{"type":58,"tag":291,"props":292,"children":294},"a",{"href":293},"references\u002Fcampaign-journal.md",[295],{"type":63,"value":293},{"type":63,"value":297},"). Keep\nit stampable: ≤400 changed lines, only files inside the allowlist below,\napply the ",{"type":58,"tag":82,"props":299,"children":301},{"className":300},[],[302],{"type":63,"value":303},"stamphog",{"type":63,"value":305}," label. Autonomy level comes from the campaign config —\ndefault is ",{"type":58,"tag":131,"props":307,"children":308},{},[309],{"type":63,"value":310},"draft PR for human review",{"type":63,"value":312},"; only arm auto-merge when the\noperator has explicitly enabled the self-driving experiment (see\nguardrails).",{"type":58,"tag":127,"props":314,"children":315},{},[316,321],{"type":58,"tag":131,"props":317,"children":318},{},[319],{"type":63,"value":320},"Journal.",{"type":63,"value":322}," Append the iteration record before ending the pass.",{"type":58,"tag":72,"props":324,"children":326},{"id":325},"hard-guardrails",[327],{"type":63,"value":328},"Hard guardrails",{"type":58,"tag":66,"props":330,"children":331},{},[332],{"type":63,"value":333},"These are not suggestions; violating any of them ends the campaign pass.",{"type":58,"tag":123,"props":335,"children":336},{},[337,401,411,421,444,454],{"type":58,"tag":127,"props":338,"children":339},{},[340,345,347,353,354,360,361,367,369,375,377,383,385,391,393,399],{"type":58,"tag":131,"props":341,"children":342},{},[343],{"type":63,"value":344},"Allowlist",{"type":63,"value":346}," — a campaign PR may only touch: ",{"type":58,"tag":82,"props":348,"children":350},{"className":349},[],[351],{"type":63,"value":352},"products\u002F*\u002Fmcp\u002Ftools.yaml",{"type":63,"value":226},{"type":58,"tag":82,"props":355,"children":357},{"className":356},[],[358],{"type":63,"value":359},"products\u002F*\u002Fskills\u002F**",{"type":63,"value":218},{"type":58,"tag":82,"props":362,"children":364},{"className":363},[],[365],{"type":63,"value":366},"services\u002Fmcp\u002Fevals\u002F**",{"type":63,"value":368},", the codegen outputs of\n",{"type":58,"tag":82,"props":370,"children":372},{"className":371},[],[373],{"type":63,"value":374},"pnpm generate-tools",{"type":63,"value":376}," \u002F ",{"type":58,"tag":82,"props":378,"children":380},{"className":379},[],[381],{"type":63,"value":382},"scaffold-yaml",{"type":63,"value":384}," (",{"type":58,"tag":82,"props":386,"children":388},{"className":387},[],[389],{"type":63,"value":390},"services\u002Fmcp\u002Fsrc\u002Ftools\u002Fgenerated\u002F**",{"type":63,"value":392},"\nand ",{"type":58,"tag":82,"props":394,"children":396},{"className":395},[],[397],{"type":63,"value":398},"services\u002Fmcp\u002Fschema\u002Fgenerated-tool-definitions.json",{"type":63,"value":400},"), and docs.\nAnything else (handler code, package manifests, workflows, migrations, auth\npaths) → stop and hand the finding to a human as a draft PR or report\ninstead.",{"type":58,"tag":127,"props":402,"children":403},{},[404,409],{"type":58,"tag":131,"props":405,"children":406},{},[407],{"type":63,"value":408},"Read-only against data.",{"type":63,"value":410}," The harness and all production queries are\nread-only. Never create, mutate, or delete customer-visible objects while\nmeasuring.",{"type":58,"tag":127,"props":412,"children":413},{},[414,419],{"type":58,"tag":131,"props":415,"children":416},{},[417],{"type":63,"value":418},"Evidence or it didn't happen.",{"type":63,"value":420}," No PR without a baseline score, an after\nscore, and the exact harness commands used.",{"type":58,"tag":127,"props":422,"children":423},{},[424,429,431,436,438,443],{"type":58,"tag":131,"props":425,"children":426},{},[427],{"type":63,"value":428},"Benchmark integrity.",{"type":63,"value":430}," Never edit ",{"type":58,"tag":82,"props":432,"children":434},{"className":433},[],[435],{"type":63,"value":95},{"type":63,"value":437}," in the same PR as\na fix it validates — changing the exam and the answer together proves\nnothing. Benchmark changes are their own PR and bump ",{"type":58,"tag":82,"props":439,"children":441},{"className":440},[],[442],{"type":63,"value":119},{"type":63,"value":121},{"type":58,"tag":127,"props":445,"children":446},{},[447,452],{"type":58,"tag":131,"props":448,"children":449},{},[450],{"type":63,"value":451},"Budgets.",{"type":63,"value":453}," Respect the operator's iteration\u002Ftoken\u002FPR caps (default: stop\nafter 3 open unmerged campaign PRs). Two failed attempts on an issue parks\nit permanently.",{"type":58,"tag":127,"props":455,"children":456},{},[457,462],{"type":58,"tag":131,"props":458,"children":459},{},[460],{"type":63,"value":461},"Kill switch.",{"type":63,"value":463}," If the campaign config, its feature flag, or the operator\nsays stop — stop mid-iteration, journal state, end cleanly.",{"type":58,"tag":72,"props":465,"children":467},{"id":466},"failure-modes-to-expect",[468],{"type":63,"value":469},"Failure modes to expect",{"type":58,"tag":123,"props":471,"children":472},{},[473,478,483],{"type":58,"tag":127,"props":474,"children":475},{},[476],{"type":63,"value":477},"A description change that helps one intent can steal traffic from the right\ntool for another — that's why the no-regression sample is mandatory.",{"type":58,"tag":127,"props":479,"children":480},{},[481],{"type":63,"value":482},"Probe latency varies with stack warmth; compare medians across ≥3 runs\nbefore attributing a latency change to your fix.",{"type":58,"tag":127,"props":484,"children":485},{},[486,488,494],{"type":63,"value":487},"Tool-presence misses can be feature-flag gating, not catalog absence —\ncheck ",{"type":58,"tag":82,"props":489,"children":491},{"className":490},[],[492],{"type":63,"value":493},"getToolsForFeatures",{"type":63,"value":495}," gating before \"fixing\" discoverability.",{"items":497,"total":593},[498,514,526,538,549,562,578],{"slug":499,"name":499,"fn":500,"description":501,"org":502,"tags":503,"stars":26,"repoUrl":27,"updatedAt":513},"analyzing-expensive-users","analyze expensive users in AI observability","Analyze the most expensive users in AI observability and explain why they cost so much. Use when the user asks about top spenders, expensive users, per-user LLM cost, user-level cost drivers, or patterns behind high AI observability spend.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[504,506,509,512],{"name":505,"slug":34,"type":15},"Analytics",{"name":507,"slug":508,"type":15},"Cost Optimization","cost-optimization",{"name":510,"slug":511,"type":15},"Observability","observability",{"name":9,"slug":8,"type":15},"2026-07-28T05:34:11.117757",{"slug":515,"name":515,"fn":516,"description":517,"org":518,"tags":519,"stars":26,"repoUrl":27,"updatedAt":525},"auditing-endpoints","audit PostHog project endpoints","Audit every endpoint in a PostHog project for staleness, failed materialisations, and unused materialised versions. Use when the user asks \"what endpoints can I clean up?\", \"are any of my endpoints broken?\", \"which materialised versions are still being called?\", or wants a one-shot cleanup pass over the Endpoints product. Produces a prioritised report grouped by issue type, with recommended actions but does not modify anything without explicit confirmation.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[520,521,524],{"name":505,"slug":34,"type":15},{"name":522,"slug":523,"type":15},"Audit","audit",{"name":9,"slug":8,"type":15},"2026-06-08T08:08:33.693989",{"slug":527,"name":527,"fn":528,"description":529,"org":530,"tags":531,"stars":26,"repoUrl":27,"updatedAt":537},"auditing-warehouse-source-health","audit PostHog data warehouse source health","Audit the health of a PostHog project's data warehouse sources and syncs — find every broken or degraded source connection, sync schema, and webhook channel. Use when the user asks \"why are my imports failing?\", \"what's broken with my sources?\", \"why is my warehouse data stale?\", or wants a one-shot triage of source\u002Fsync health before deciding where to dig in. Produces a prioritized report grouped by severity, with recommended next steps. For materialized-view health use `auditing-warehouse-view-health`; for a single failing sync use `diagnosing-failed-warehouse-syncs`.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[532,533,535,536],{"name":522,"slug":523,"type":15},{"name":534,"slug":36,"type":15},"Data Warehouse",{"name":510,"slug":511,"type":15},{"name":9,"slug":8,"type":15},"2026-06-18T08:22:57.67984",{"slug":539,"name":539,"fn":540,"description":541,"org":542,"tags":543,"stars":26,"repoUrl":27,"updatedAt":548},"auditing-warehouse-view-health","audit PostHog materialized view health","Audit the health of a PostHog project's materialized views (saved queries) — find every failed materialization and flag unused or stale materialized views that cost storage and compute. Use when the user asks \"which of my views are broken?\", \"why is this materialized view failing?\", \"are any of my views wasting compute?\", or wants a one-shot triage of view health. For source\u002Fsync health use `auditing-warehouse-source-health`.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[544,545,546,547],{"name":522,"slug":523,"type":15},{"name":534,"slug":36,"type":15},{"name":13,"slug":14,"type":15},{"name":9,"slug":8,"type":15},"2026-06-18T08:25:10.936787",{"slug":550,"name":550,"fn":551,"description":552,"org":553,"tags":554,"stars":26,"repoUrl":27,"updatedAt":561},"authoring-error-tracking-alerts","author PostHog error tracking alerts","Author error tracking alerts that fire when an issue is created, reopened, or starts spiking. Use when the user asks to set up error notifications, route exceptions to Slack\u002Fwebhook\u002FLinear, or evaluate which error events are worth alerting on. Covers trigger-event selection, integration choice, dedup against existing alerts, and shipping with the canonical message body shape.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[555,558,559,560],{"name":556,"slug":557,"type":15},"Alerting","alerting",{"name":24,"slug":25,"type":15},{"name":510,"slug":511,"type":15},{"name":9,"slug":8,"type":15},"2026-06-18T08:24:40.318583",{"slug":563,"name":563,"fn":564,"description":565,"org":566,"tags":567,"stars":26,"repoUrl":27,"updatedAt":577},"authoring-log-alerts","author log alerts in PostHog","Author useful, low-noise log alerts on services in a PostHog project. Use when the user asks to set up alerts for their logs, suggest alerts they should add, or evaluate whether a service is worth monitoring. Covers service triage, baseline characterisation, threshold drafting, back-testing via simulate, and shipping with a notification destination.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[568,569,572,573,576],{"name":505,"slug":34,"type":15},{"name":570,"slug":571,"type":15},"Monitoring","monitoring",{"name":510,"slug":511,"type":15},{"name":574,"slug":575,"type":15},"Operations","operations",{"name":9,"slug":8,"type":15},"2026-07-18T05:10:54.430898",{"slug":579,"name":579,"fn":580,"description":581,"org":582,"tags":583,"stars":26,"repoUrl":27,"updatedAt":592},"building-workflows","build and edit PostHog workflows","Build, edit, test, enable, and monitor PostHog workflows over MCP. Author the action\u002Fedge graph so it runs and opens cleanly in the visual editor, then change drafts surgically with patch operations. Use when asked to build, set up, automate, change, fix, or debug a workflow, campaign, broadcast, drip sequence, or event-triggered automation in the workflows product.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[584,587,588,589],{"name":585,"slug":586,"type":15},"Automation","automation",{"name":21,"slug":22,"type":15},{"name":9,"slug":8,"type":15},{"name":590,"slug":591,"type":15},"Workflow Automation","workflow-automation","2026-07-28T05:34:12.167015",61,{"items":595,"total":712},[596,603,609,616,623,630,638,645,659,674,684,702],{"slug":499,"name":499,"fn":500,"description":501,"org":597,"tags":598,"stars":26,"repoUrl":27,"updatedAt":513},{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[599,600,601,602],{"name":505,"slug":34,"type":15},{"name":507,"slug":508,"type":15},{"name":510,"slug":511,"type":15},{"name":9,"slug":8,"type":15},{"slug":515,"name":515,"fn":516,"description":517,"org":604,"tags":605,"stars":26,"repoUrl":27,"updatedAt":525},{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[606,607,608],{"name":505,"slug":34,"type":15},{"name":522,"slug":523,"type":15},{"name":9,"slug":8,"type":15},{"slug":527,"name":527,"fn":528,"description":529,"org":610,"tags":611,"stars":26,"repoUrl":27,"updatedAt":537},{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[612,613,614,615],{"name":522,"slug":523,"type":15},{"name":534,"slug":36,"type":15},{"name":510,"slug":511,"type":15},{"name":9,"slug":8,"type":15},{"slug":539,"name":539,"fn":540,"description":541,"org":617,"tags":618,"stars":26,"repoUrl":27,"updatedAt":548},{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[619,620,621,622],{"name":522,"slug":523,"type":15},{"name":534,"slug":36,"type":15},{"name":13,"slug":14,"type":15},{"name":9,"slug":8,"type":15},{"slug":550,"name":550,"fn":551,"description":552,"org":624,"tags":625,"stars":26,"repoUrl":27,"updatedAt":561},{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[626,627,628,629],{"name":556,"slug":557,"type":15},{"name":24,"slug":25,"type":15},{"name":510,"slug":511,"type":15},{"name":9,"slug":8,"type":15},{"slug":563,"name":563,"fn":564,"description":565,"org":631,"tags":632,"stars":26,"repoUrl":27,"updatedAt":577},{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[633,634,635,636,637],{"name":505,"slug":34,"type":15},{"name":570,"slug":571,"type":15},{"name":510,"slug":511,"type":15},{"name":574,"slug":575,"type":15},{"name":9,"slug":8,"type":15},{"slug":579,"name":579,"fn":580,"description":581,"org":639,"tags":640,"stars":26,"repoUrl":27,"updatedAt":592},{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[641,642,643,644],{"name":585,"slug":586,"type":15},{"name":21,"slug":22,"type":15},{"name":9,"slug":8,"type":15},{"name":590,"slug":591,"type":15},{"slug":646,"name":646,"fn":647,"description":648,"org":649,"tags":650,"stars":26,"repoUrl":27,"updatedAt":658},"check-posthog-loading","inspect PostHog SDK loading across URLs","Inspect how the PostHog JavaScript SDK is loaded across a list of URLs. Use to confirm consistent installation across pages, find pages missing the snippet, detect mismatched API keys or hosts between pages, and verify the load method (head snippet vs deferred vs array.js).\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[651,652,653,656,657],{"name":505,"slug":34,"type":15},{"name":24,"slug":25,"type":15},{"name":654,"slug":655,"type":15},"Frontend","frontend",{"name":510,"slug":511,"type":15},{"name":9,"slug":8,"type":15},"2026-05-07T05:56:19.828048",{"slug":660,"name":660,"fn":661,"description":662,"org":663,"tags":664,"stars":26,"repoUrl":27,"updatedAt":673},"consuming-endpoints-from-client-code","integrate PostHog endpoints into client applications","Wire a PostHog endpoint into a client app or SDK. Covers fetching the OpenAPI spec, generating a typed client with openapi-generator or @hey-api\u002Fopenapi-ts, sending the right auth header, shaping the variables payload (HogQL code_name vs insight breakdown property), handling rate-limit and materialised-endpoint error responses. Use when the user says \"how do I call my endpoint\", \"generate a client for this\", or \"what auth header do I use\".\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[665,668,669,670],{"name":666,"slug":667,"type":15},"API Development","api-development",{"name":654,"slug":655,"type":15},{"name":9,"slug":8,"type":15},{"name":671,"slug":672,"type":15},"SDK","sdk","2026-06-08T08:08:34.929454",{"slug":675,"name":675,"fn":676,"description":677,"org":678,"tags":679,"stars":26,"repoUrl":27,"updatedAt":683},"copying-endpoints-across-projects","copy PostHog endpoints across projects","Copy a PostHog endpoint (a saved HogQL\u002Finsight query exposed as an API route) to another project in the same organization, or duplicate it under a new name in the same project. Use when the user wants to duplicate an endpoint, promote an endpoint from staging to production, replicate an endpoint's query\u002Fvariables\u002Ffreshness config in another workspace, or clone an endpoint to iterate on it. Unlike feature flags and experiments, endpoints have NO native cross-project copy tool — this skill covers the read-then-recreate flow (endpoint-get then endpoint-create), the active-project switching it requires, name-collision checks, and the safe defaults (land unmaterialised in the target, verify with endpoint-run). Does not cover editing endpoint versions (see managing-endpoint-versions) or authoring a brand-new endpoint from scratch (see creating-an-endpoint).\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[680,681,682],{"name":666,"slug":667,"type":15},{"name":574,"slug":575,"type":15},{"name":9,"slug":8,"type":15},"2026-07-15T05:29:58.442727",{"slug":685,"name":685,"fn":686,"description":687,"org":688,"tags":689,"stars":26,"repoUrl":27,"updatedAt":701},"creating-ai-subscription","schedule recurring AI-generated PostHog reports","Create a recurring AI-generated PostHog report — schedule a free-text prompt to run on a cron, with the LLM-synthesized markdown delivered to email or Slack on each tick. Use when the user wants a recurring AI summary of X on any cadence (daily, weekly, monthly, yearly) rather than a one-off report. (To attach an AI summary to an existing insight\u002Fdashboard subscription instead of a free-text prompt, see `managing-subscriptions` and its `summary_enabled` option.)\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[690,691,694,695,698],{"name":585,"slug":586,"type":15},{"name":692,"slug":693,"type":15},"Email","email",{"name":9,"slug":8,"type":15},{"name":696,"slug":697,"type":15},"Reporting","reporting",{"name":699,"slug":700,"type":15},"Slack","slack","2026-06-09T07:32:27.935712",{"slug":703,"name":703,"fn":704,"description":705,"org":706,"tags":707,"stars":26,"repoUrl":27,"updatedAt":711},"creating-an-endpoint","create PostHog API endpoints","Create a PostHog endpoint with the right shape on the first try — covers query kind choice, name conventions, what to expose as variables (HogQL code_name vs insight breakdown), data_freshness_seconds, and whether to materialise on day one. Use when the user says \"create an endpoint\", \"expose this query as an API\", \"turn this insight into an endpoint\", or asks for help structuring a new endpoint. Steers away from common mistakes: materialising a query with cohort breakdowns or compare mode, inline-only variables on a materialised endpoint, unbounded date ranges, ambiguous names.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[708,709,710],{"name":505,"slug":34,"type":15},{"name":666,"slug":667,"type":15},{"name":9,"slug":8,"type":15},"2026-06-08T08:08:29.624498",231]