Agent Claude Code
Assesses whether it is safe to approve the gated CI workflows on an outside-contributor (fork) pull request. GitHub holds a fork PR's workflows until a maintainer clicks "Approve and run", because approval lets the contributor's code execute in CI with access to the base repo's runners and secrets. This agent reads…
Agent Claude Code
Use this agent when you encounter a failing test, unexpected behavior, or need to investigate the underlying cause of a software defect. Examples: Context: User is debugging a failing integration test that worked yesterday but now fails intermittently. user: "This test keeps failing randomly - sometimes it passes…
Agent Claude Code
Reviews code for bugs, logic errors, security vulnerabilities, and project guideline violations. Does not cover readability or refactoring (use the simplify skill for that). Examples: before committing changes, after implementing a new feature, or when you want a correctness check.
Agent Claude Code
Use this agent when you need to break down complex software features or requirements into actionable implementation stages, create technical specifications, or design system architecture before coding begins. Examples: Context: User wants to add a new authentication system to their web application. user: 'I need to…
Agent Claude Code
Investigates a single operational hypothesis using production metrics, logs, dashboards, and PostHog product data (Grafana/Prometheus/Loki/PostHog DB/PostHog data warehouse). Spawn one instance per hypothesis so the orchestrator can fan out in parallel. Use when the question is answered by observability data, not…
Agent Claude Code
Use this agent when you need to improve, refine, or optimize prompts for AI systems, LLMs, or other automated tools. Examples include: refining system prompts for better performance, analyzing prompt effectiveness, restructuring prompts for clarity and precision, optimizing prompts for specific use cases or domains…
Agent Claude Code
Triages a single CI flake (a failing test or a flaky infra step) and, if it looks like a genuine unknown flake, reports it to the @PostHog bot in the #flakey-tests Slack channel so the bot can investigate. Fire-and-forget: spawn it with a failing job URL and keep working. It does NOT root-cause, fix, or reproduce the…
Agent Claude Code
Analyzes GitHub issues and PRs (external and internal) to identify those belonging to the Feature Flags team domain (feature flags, cohorts, early access). Used by the triage-issues command.
Agent Claude Code
Use this agent when you need to write comprehensive unit tests for existing code, when implementing test-driven development, when code coverage needs improvement, or when refactoring requires test safety nets. Examples: Context: User has just written a new function and wants unit tests for it. user: 'I just wrote this…
Skill Claude CodeCodex
Simplify recently changed code for clarity, consistency, and maintainability while preserving behavior. Use after implementation or when asked to simplify or refactor code without changing what it does.
Skill Claude Code
Evaluate unresolved PR review comments from any reviewer — bots and humans — fix legitimate issues, and reply to dismissed ones.
Skill Claude Code
Analyze accumulated permissions and suggest smart wildcard patterns. Only invoke when the user explicitly runs /analyze-permissions or asks to analyze their Claude Code permissions.
Skill Claude Code
One sweep over all of my open PRs — check CI and merge-queue state, handle new review comments, fix and push — tracking state so reruns skip already-handled work. Designed to be driven by /loop.
Skill Claude Code
Monitor CI checks after pushing, detect flaky vs legit failures, and auto-fix. Follows a PR into a Trunk merge queue, where the queue's CI runs on its own branch.
Skill Claude Code
Delete and tighten code comments in source files after they are written. Drops narration, restatement, and change-log comments, cuts the survivors to their essential facts, and rewrites the grammar of what remains. Use when the user says the comments are too wordy, asks to clean up or trim comments, or right after…
Skill Claude Code
Commit staged/unstaged changes with a well-crafted commit message.
Skill Claude Code
Create or update a GitHub PR with automatic template detection and filling, enforcing a plain staff-engineer writing voice in the PR body.
Skill Claude Code
Explain each open or skipped code-review item in plain English, weigh what happens on each side of the decision, and give a recommendation.
Skill Claude Code
Capture a follow-up item mid-session in seconds, list open items, close one, or run a review pass. Use when the user says "follow up on this later", "add a followup", "don't let me forget", or runs /followup.
Skill Claude CodeCodex
Reference for GitHub PR review endpoints and resolving review threads via gh CLI. Use when replying to a PR review comment, posting a new review comment, or resolving review threads.
Skill Claude Code
Plan, implement, and review a task end-to-end — review-code + ReviewHog in parallel, every review addressed, PR marked ready, CI watched to green, stamphog asked to approve, open items explained. Idempotent — re-running reports where the pipeline stands and resumes from the first incomplete step.
Skill Claude Code
Write or resume a handoff document so the next agent session can pick up the current work. Use when context is filling up, when ending a session mid-task, or at the start of a new session that should continue prior work.
Skill Claude Code
Query PostHog production Metabase. Use when the user wants prod data such as counts, distributions, customer-report debugging, or region comparisons. Not for local dev investigations; use the posthog-db MCP for those.
Skill Claude Code
Capture complex technical discoveries into structured, reusable notes. Use when explaining system behaviors, documenting debugging insights, or preserving knowledge.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: