Diagnoses and fixes React codebase health issues, including performance problems, security gaps and code quality.
PostHog/posthog
Guides for working inside the PostHog codebase, covering how to add features, run tests, and ship changes safely.
- 200 skills
- 41k stars
The 5 most installed
Ordered by installs on skills.sh- react-doctor511 installs
- hogli267 installs
Explains the PostHog developer CLI and repo tooling for running, building, testing, and linting.
- clickhouse-migrations266 installs
Writes and modifies ClickHouse database migrations following project rules.
- survey-sdk-audit255 installs
Audits survey SDK features and version requirements, then implements survey changes across backend, UI, and SDKs.
- django-migrations252 installs
Guides safe Django and Postgres migration work, including non-blocking schema changes, backfills and model deletions.
The other 195 skills
- setup-web-tests252 installs
Sets up a Python test environment in Claude Code for web when the usual sync command fails.
- ingestion-pipeline-doctor-nodejs225 installs
Explains the ingestion pipeline architecture and points to the right doctor agent for a given concern.
- playwright-test225 installs
Writes a Playwright test, makes sure it runs, and checks that it is not flaky.
- implementing-mcp-tools212 installs
Exposes a PostHog API endpoint as an MCP tool, from serializer through generated TypeScript handler.
- improving-drf-endpoints209 installs
Reviews and improves Django REST Framework endpoints and serializers, covering field typing, schema annotations, and OpenAPI quality.
- writing-skills205 installs
Guides writing PostHog agent skills that teach agents how to use MCP tools to reach a goal.
- implementing-warehouse-sources194 installs
Adds or extends data warehouse import sources, including datasets, incremental sync, pagination, credentials, and tests.
- isolating-product-facade-contracts192 installs
Plans and executes product isolation migrations toward a facade plus contract layer with a low-conflict PR strategy.
- qa-team191 installs
Runs a multi-agent code review of your branch, with specialists covering security, database, performance, frontend and copy.
- adopting-generated-api-types181 installs
Migrates frontend code from hand-written API calls and types to generated API functions.
- modifying-taxonomic-filter179 installs
Guides safe changes to PostHog's event and property picker, covering search, ordering, recents and performance.
- sending-notifications179 installs
Shows how to send real-time in-app notifications from backend code, including target type and priority.
- implementing-mcp-ui-apps176 installs
Adds interactive MCP UI views that render tool results inside clients like Claude Desktop.
- security-audit171 installs
Builds a threat model from a diff or PR and checks the code for access control, injection, and data leaks.
- debugging-ci-failures163 installs
Investigates failing GitHub Actions CI runs and answers questions about overall CI and master branch health.
- monitoring-capture-service162 installs
Checks the health of PostHog's capture service in production, covering latency, event loss and Kafka backpressure.
- monitoring-ingestion-pipeline160 installs
Diagnoses PostHog's ingestion workers in production, covering event lag, drops, pipeline errors and downstream health.
- adding-personhog-rpc159 installs
Adds a new gRPC endpoint to the personhog service, including proto, clients and Rust code.
- writing-tests153 installs
Decides whether a new test is worth adding and keeps pytest, Jest and Playwright tests cheap and meaningful.
- adding-project-secret-api-key-auth148 installs
Lets an API endpoint accept project secret API keys for programmatic, user-less access.
- fixing-flaky-tests148 installs
Reproduces, root-causes, fixes, and validates intermittently failing Jest, pytest, or Playwright tests.
- optimizing-clickhouse-and-hogql-queries148 installs
Speeds up slow ClickHouse and HogQL queries and explains how to fix timeouts or memory limits.
- run-posthog147 installs
Starts, inspects, and drives the local PostHog dev stack to check health or verify a UI change.
- writing-code-comments147 installs
Decides whether a code comment should exist and keeps only ones that explain why, not what.
- managing-github-actions-secrets146 installs
Creates and updates GitHub Actions secrets for PostHog workflows, including CI tokens, registry credentials, and deploy keys.
- setting-up-devbox146 installs
Starts, connects to, and troubleshoots a remote PostHog devbox workspace through command-line tools.
- using-kea-disposables146 installs
Shows how to register timers and event listeners in kea logic so they are cleaned up automatically.
- writing-clickhouse-queries146 installs
Guides writing fast ClickHouse queries, choosing workloads, table design, materialized columns and indexes.
- depot-github-runners145 installs
Sets up Depot-managed GitHub Actions runners, including sizes, labels and ARM or Windows runners.
- manage-dashboard-widgets145 installs
Guides shipping a new dashboard widget type or updating a shipped one, including config, queries, layout, and RBAC.
- writing-kea-logics145 installs
Guides writing and reviewing kea logic files, including actions, reducers, selectors, listeners and React wiring.
- depot-container-builds144 installs
Builds Docker images with Depot, including multi-platform builds and layer caching fixes.
- django-startup-time144 installs
Keeps heavy imports off the Django startup path so every process starts faster.
- generating-clickhouse-query-performance-reports142 installs
Produces slow-query performance reports for production ClickHouse, covering cost, timeouts, and memory regressions.
- visualizing-change-over-time142 installs
Helps choose the right chart for before/after and period-over-period comparisons when building UI with @posthog/quill-charts.
- documenting-warehouse-sources141 installs
Writes or updates the user-facing documentation page for a data warehouse import source.
- establishing-code-ownership141 installs
Finds which team owns a file or code path, or lists everything a given team owns.
- gating-production-deploys141 installs
Adds or edits GitHub Actions workflows that push images or trigger production deploys, with the required gate.
- querying-local-postgres141 installs
Runs read-only SQL against the local Postgres database to inspect tables, debug data and analyze query plans.
- running-ci-preflight132 installs
Runs local preflight checks to catch formatting, lint, lockfile, and migration failures before pushing.
- integrating-with-posthog-ai83 installs
Wires a PostHog frontend surface into the PostHog AI agent with scene context, instructions, tool reactions, and tool cards.
- writing-user-facing-copy30 installs
Sets the voice and word choices for any text users read in PostHog, from buttons to error messages and docs.
- adding-ingestion-warnings24 installs
Adds a new warning type to the event ingestion pipeline with the right category and severity.
- analyzing-experiment-query-performance24 installs
Pulls production experiment query-performance data and explains slow or failing experiment queries.
- qa-frontend24 installs
Runs browser-based QA on PostHog frontend changes and can record a feature demo reel.
- querying-tophog24 installs
Queries the ingestion pipeline's ClickHouse store to find which teams or users are causing lag or high cost.
- writing-pr-descriptions24 installs
Shapes a pull request description so reviewers quickly understand the problem and the change.
- adding-inbox-sources23 installs
Connects a data warehouse table as a source of actionable items for the Self-driving Inbox.
- building-product-empty-states23 installs
Adds first-run setup empty states to product screens using the shared component.
- stacking-prs23 installs
Creates and manages GitHub stacked pull requests with the gh stack tool, including rebasing and landing.
- adding-mcp-store-servers22 installs
Adds a third-party MCP server to the PostHog MCP store catalog and verifies it works.
- adding-product-alerting22 installs
Adds alerting to a PostHog product or extends the shared alerts platform.
- debugging-local-task-agent-runs22 installs
Helps read logs and diagnose stuck, failed or silent local task agent runs.
- extending-personhog-test-harness22 installs
Adds regression scenarios, chaos events, and invariants to the personhog end-to-end test harness.
- instrumenting-first-party-metrics22 installs
Records counters, gauges, and histograms into PostHog's own metrics product from PostHog code.
- writing-evals22 installs
Teaches how to write and run eval suites, cases, scorers and seeders on the PostHog AI eval harness.
- analyzing-insights-across-teams21 installs
Analyzes insights and dashboards across all teams by querying the synced dogfood warehouse.
- warehouse-source-new-version21 installs
Adds support for a new vendor API version to an existing data warehouse source, or deprecates an old one.
- auditing-warehouse-source-coverage20 installs
Finds vendor API endpoints and tables a data warehouse source never wired up, and ranks what to add next.
- triaging-web-analytics-support19 installs
Sorts web analytics support tickets into known problem types, runs the matching playbook and drafts replies or fixes.
- merging-prs18 installs
Merges a PR through the Trunk merge queue and watches it until it lands or gets kicked out.
- posthog-desktop17 installs
Scopes work to the PostHog desktop Electron app and its separate workspace, packages and agent framework.
- reviewing-personhog-protocol17 installs
Runs a thorough two-pass review of personhog coordination-protocol changes covering leases, fencing, handoffs and failure budgets.
- using-redis-token-buckets17 installs
Guides adding Redis-backed rate limits with burst capacity, refills and refunds for requests that did no work.
- auditing-llm-gateway-parity16 installs
Compares the two LLM gateways and updates the parity record from current implementation evidence.
- finding-llm-gateway-migration-candidates16 installs
Ranks callers that could safely move from the old LLM gateway to the new one, with evidence.
- managing-dashboards16 installs
Guides dashboard platform and scene changes, including layouts, sharing, embeds, filters, refresh behavior, and performance.
- migrating-llm-gateway-callers16 installs
Migrates an LLM caller from the old gateway to PostHog ai-gateway, updating clients, URLs, headers, and tests.
- placing-product-frontend-code16 installs
Decides which frontend folder a product file belongs in and explains the reasoning behind the split.
- splitting-oversized-modules16 installs
Splits a very large Python module into smaller per-concern modules without changing behavior.
- triaging-warehouse-sync-tickets16 installs
Checks the real sync state of a customer's warehouse table in production to debug stale, empty or failing imports.
- debugging-table-access-denied15 installs
Investigates a table access denied error to tell whether it is a real bug or expected behavior.
- extending-hobby-smoke-tests15 installs
Adds, reviews, or debugs PostHog Hobby end-to-end smoke tests and their CI workflow.
- writing-dataclasses15 installs
Applies house rules for Python dataclasses: when to use one, naming, construction and keeping secrets out of repr.
- writing-ui-components15 installs
Sets structure and abstraction rules for React components and frontend files, including file layout and shared components.
- improving-mcp-tools14 installs
Runs an eval-backed campaign that measures MCP tool quality, fixes the highest-impact problem, and keeps changes only if scores improve.
- announcing-behavior-changes13 installs
Decides whether a behavior-changing fix needs an in-app notice and builds one for affected users only.
- debugging-local-replay13 installs
Traces why session recordings don't show up in the local dev environment and points to the broken step.
- maintaining-python-tests13 installs
Maintains existing pytest and Django test suites, reducing runtime and stale tests without weakening correctness.
- debugging-signals-pipeline12 installs
Traces the signals pipeline locally to find why a signal never reaches the inbox.
- triaging-merge-queue-failures12 installs
Explains why a pull request was kicked from the merge queue and tells you whether to wait, fix, requeue or escalate.
- adding-activity-logging11 installs
Adds or fixes the audit trail that records who changed a Django model and when.
- analyzing-experiment-precompute-canary10 installs
Checks the experiment precompute canary for divergences and pinpoints the affected team, experiment and metric.
- debugging-mcp-analytics10 installs
Debugs missing or wrong MCP analytics data and helps build on the MCP analytics SDKs.
- depot-ci10 installs
Sets up and manages Depot CI, including migrating GitHub Actions workflows and debugging runs.
- editing-agents-md10 installs
Decides whether a rule belongs in AGENTS.md or CLAUDE.md and writes it so it is followed.
- organizing-conversations-code10 installs
Explains where files for the conversations product belong when adding, moving or reviewing them.
- profiling-slow-api-endpoints10 installs
Profiles slow PostHog API endpoints whose cost sits in Postgres or Python, and plans fixes with production evidence.
- querying-production-databases-via-metabase10 installs
Queries production databases through PostHog's internal Metabase for query plans, index checks and shareable links.
- reviewing-with-coderabbit10 installs
Does nothing, because CodeRabbit is retired on this repository in favor of PostHog Review.
- setting-up-support-slack-locally10 installs
Connects a real Slack workspace to local PostHog Conversations so Slack messages become support tickets.
- testing-mcp-tools-locally10 installs
Sets up local data and API keys to test staff-only managed migrations MCP tools end to end.
- working-with-charts10 installs
Guides chart selection, data, themes, sizing and composition when building charts with @posthog/quill-charts.
- adding-inbound-webhooks9 installs
Adds an endpoint that receives webhooks from a third-party service and processes them safely.
- authenticating-to-clickhouse9 installs
Connects a new or changed service to ClickHouse using rotating service-account tokens instead of static passwords.
- developing-subscriptions9 installs
Guides changes to PostHog subscriptions, covering schedules, delivery, retries, permissions and related UI.
- routing-outbound-api-calls9 installs
Routes new third-party HTTP calls through the right PostHog egress domain and lists the tests that catch mistakes.
- gating-sensitive-actions8 installs
Decides when an endpoint, setting, or page should require recent re-authentication and wires it up.
- adding-api-scopes7 installs
Adds a new API scope so personal keys, OAuth tokens and MCP clients can access an endpoint.
- exploring-llm-traces5 installs
Inspects a pasted LLM trace or session to show what the agent did, which tools it used, and what it cost.
- analyzing-experiment-session-replays4 installs
Compares session replays across experiment variants to show how user behavior differs.
- cleaning-up-stale-feature-flags4 installs
Finds unused or fully rolled-out feature flags and removes the code that checks them.
- configuring-experiment-analytics4 installs
Sets up exposure events and metrics for a PostHog experiment and explains how to read results.
- configuring-experiment-rollout4 installs
Sets the variant split and rollout percentage for a PostHog experiment.
- creating-experiments4 installs
Walks through setting up a new A/B test, from hypothesis and rollout to metrics and run time.
- diagnosing-sdk-health4 installs
Checks which PostHog SDKs in a project are out of date and how to upgrade them.
- exploring-llm-clusters4 installs
Investigates AI traffic clusters, compares their behavior, and drills into individual traces.
- exploring-llm-evaluations4 installs
Finds, inspects, and runs AI evaluations, then reports on their results and failure modes.
- feature-usage-feed4 installs
Sets up an LLM-judge evaluation that extracts real use cases for a feature and streams them to Slack.
- finding-experiments4 installs
Turns a loose description of an experiment into the exact experiment ID, asking when several match.
- managing-experiment-lifecycle4 installs
Guides experiment state transitions like launching, pausing, ending, shipping variants, archiving, duplicating, and migrating legacy experiments.
- copying-flags-across-projects3 installs
Copies a feature flag to other PostHog projects, remapping cohorts and handling encrypted payloads.
- exploring-llm-costs3 installs
Breaks down LLM spending by model, provider, user, or feature and explains cost spikes.
- auditing-endpoints2 installs
Audits a project's endpoints for staleness, failed materialisations and unused versions, then reports what to clean up.
- auditing-warehouse-view-health2 installs
Finds failed, unused or stale materialized views that waste storage and compute in a project.
- consuming-endpoints-from-client-code2 installs
Wires a PostHog endpoint into an app by generating a typed client and handling auth and errors.
- creating-an-endpoint2 installs
Creates a PostHog endpoint that exposes a query as an API with the right shape.
- creating-online-evaluations2 installs
Creates continuously running evaluations that automatically score new AI responses and traces for known failures.
- diagnosing-ci-and-merge-bottlenecks2 installs
Uses engineering analytics tools to find slow CI workflows, stuck pull requests and costly failing tests.
- diagnosing-endpoint-performance2 installs
Explains why a single PostHog endpoint is slow or costly and proposes a concrete fix.
- diagnosing-stacktrace-symbolication2 installs
Helps fix error stack traces that stay minified after symbols are uploaded, across web, mobile and desktop.
- exploring-ai-failures2 installs
Finds where an AI feature is failing in production and groups the failure patterns from real traces.
- exploring-endpoint-execution-logs2 installs
Reads a PostHog endpoint's execution logs to explain errors, failed runs, cache misses, and slow runs.
- finding-deleted-feature-flags2 installs
Finds feature flags deleted in the active project recently, including who deleted them and when.
- grouping-noisy-errors2 installs
Merges error tracking issues that are the same error under different fingerprints, or adds a lasting grouping rule.
- investigating-error-issue2 installs
Investigates one error tracking issue end-to-end, covering affected users, triggers, timing, correlations, and next steps.
- managing-endpoint-versions2 installs
Previews drafts, rolls back, updates settings, or deactivates specific endpoint versions and explains current limitations.
- suppressing-noisy-errors2 installs
Creates error tracking suppression rules to drop high-volume, low-value errors at ingestion.
- triaging-error-issues2 installs
Triages error tracking issues into a prioritized list, ranking by users affected and pointing at linked replays.
Adds a new highlight rule that surfaces a product's facts on the Today home page.
Syncs warehouse table columns onto person or group properties for flags, cohorts and filters.
Answers whether AI answer engines cite your site, which topics get cited least, and who gets cited instead.
Finds the users driving the most AI observability cost and explains what makes them expensive.
Compares offline AI evaluation runs and diagnoses why individual cases regressed.
Creates a new dashboard or updates an existing one by assembling insights into one page.
Builds or edits a PostHog freeform canvas app such as a data board, form, or small tool.
Builds PostHog canvases using plain HTML, CSS, and browser APIs for documents and graphics.
Guides sending events to PostHog's capture v1 batch intake API and changing its schema.
Builds PostHog canvases with React and the Quill design system so they look native.
Builds, watches, and tests the PostHog toolbar and its browser extension.
Explains what a member or role can view or edit in a PostHog project.
Adds, moves, resizes, or removes widgets on PostHog grid and home canvases.
Keeps the context wiki coherent by working through a lint report queue.
Synthesizes recent organizational activity into a durable context wiki.
Reviews the context wiki for contradictions, stale priorities, and duplicate subjects.
Copies or duplicates a PostHog endpoint into another project or under a new name.
Adds or changes widgets, tiles and tabs on the customer analytics account detail page.
Diagnoses why a customer's A/B test results look wrong, such as uneven exposures or mismatched numbers.
Finds why a data quality check failed, shows the failing rows and recommends a fix.
Guides changes to PostHog's two command line tools, keeping the Rust and TypeScript versions in step.
Runs a health check on one experiment, covering exposures, sample ratio mismatch, significance traps and mid-run edits.
Exports PostHog events, persons, sessions or query results and downloads the resulting files.
Runs large audits, investigations, and triage as many isolated agents coordinated by a JavaScript workflow.
Adds panels, staff actions, and cross-team reports to the PostHog Admin staff console.
Builds and validates Slack Block Kit messages, modals, and App Home layouts.
Investigates a CI failure to a verdict: whose fault, which commit, who wrote it, and whether a fix landed.
Links an error occurrence or session to OpenTelemetry traces, same-trace logs, and Session Replay.
Creates, versions, releases, and fetches LLM prompts with PostHog prompt management and SDK code.
Installs, configures, authenticates, and troubleshoots MCP servers, including OAuth login and mcp.json setup.
Builds reusable activation models, including an activation-rate metric and a per-user activated flag, on warehouse views or dbt.
Builds reusable conversion models such as funnel rates, drop-off, and time-to-convert on warehouse views or dbt.
Builds reusable dimension and lookup tables for a star schema, like country, timezone, currency, date, and plan.
Builds reusable product-usage models for retention, stickiness, and lifecycle on warehouse views or dbt.
Builds reusable revenue models like MRR, ARR, churn, ARPU, LTV, and per-customer revenue on warehouse views or dbt.
Sets up reusable data models in PostHog's warehouse or a dbt project before you build business models.
Shows how to pull PostHog insights and queries into a canvas correctly, including date ranges and live third-party data.
Runs read-only queries against PostHog's production ClickHouse and Postgres in the US and EU regions.
Edits the @posthog/quill design system locally and tests the change in the desktop app before publishing.
Records a captioned, branded walkthrough video of a feature running on the local PostHog stack.
Re-creates a reviewed pull request from a fork on a PostHog branch so it can pass checks and merge.
Decides whether a mobile app change ships as an over-the-air update or needs a native store release, then files the request.
Diagnoses ingestion warnings such as missing events, failed merges, oversized payloads and empty heatmaps.
Runs the mobile app's demo smoke flows on a local iOS simulator and reports their results.
Sets up a Replay Vision scanner that watches an experiment's sessions and compares results across variants.
Sets up AI citation tracking for a project, choosing answer engines, competitors, and prompts to monitor.
Populates a project's data catalog with canonical metrics, trust marks, and reviewed table relationships.
Maps warehouse table columns to person or group properties usable in flags, cohorts, and insights.
Watches AI answer engine citation rates and files a report when a domain's rate shifts or the checker fails.
Writes Storybook stories for UI components, covering provider setup and data-fetching pitfalls.
Decides when to delegate codebase exploration, planning, or code changes to subagents with separate context.
Drives the running PostHog desktop app over CDP to test changes against a live local backend.
Turns PR and CI data into saved PostHog insights, dashboards and scheduled reports, with SQL examples.
Explains what is driving PostHog usage and spend, and how to reduce it.
Syncs the forked Claude agent adapter with a newer upstream release while keeping local changes.
Validates and publishes a canvas project safely, handling version conflicts and capacity limits.
Picks the cheapest check that confirms a change works and says when a slower one is needed.
Trains agent-driven models that predict who will churn, convert or adopt a feature, then scores people daily.
Runs, writes and debugs Storybook stories and their visual review snapshots.