Claude Skills for developers
PR descriptions, code review, test generation, postmortems, ADRs, on-call runbooks, migration plans. The engineering workflow as installable Claude skills.
How to install
Claude Code (CLI): Download the pack, then unzip into your skills folder.
mkdir -p ~/.claude/skills
unzip ~/Downloads/developer-pack.zip -d ~/.claude/skills/For project-only skills, use .claude/skills/ in your repo root instead. Restart Claude Code (or the session) and the skills become available.
Claude.ai (web): Upload via Settings → Skills (where supported on your plan).
Skills in this pack
PR Description Writer
Generates a PR description from a diff — what changed, why, how to test, risks.
Code Review Helper
Reviews a diff for bugs, edge cases, naming, and test coverage. Direct, not nitpicky.
Test Generator
Generates unit/integration tests for a function or module — happy path, edge cases, failures.
ADR Writer
Writes an Architecture Decision Record — context, decision, alternatives, consequences.
Postmortem Writer
Writes a blameless postmortem from incident notes — timeline, root cause, action items.
On-call Runbook Builder
Builds a runbook for a specific alert — symptoms, diagnosis, remediation, escalation.
Migration Plan Drafter
Drafts a phased migration plan — risks, rollback, dual-run period. No big-bang switches.
Refactor Planner
Plans a refactor as small reversible steps — what changes per step, what to test between.
Bug Repro Builder
Builds a minimal reproducible example from a bug report — strips noise, isolates the failure.
API Documentation Writer
Documents an API endpoint or library function — examples, errors, edge cases.
Commit Message Writer
Generates conventional commit messages from staged changes — subject + body, why-not-what.
Release Notes Writer
Generates user-facing release notes from merged PRs — grouped, highlighted, breaking changes flagged.
Debug Helper
Walks a stack trace + repro into a root-cause hypothesis. Asks for missing context before guessing.
AI Output Verifier
Audits AI-generated code for hallucinated APIs, made-up signatures, and silent rewrites.
Loop Breaker
Diagnoses why an AI agent is stuck and produces a clean restart with the right context.
Context Grounding Brief
Builds a context bundle (real schema, real types, real signatures) before asking AI to write code.