claude-skills

Expert guidance for ffuf web fuzzing during penetration testing, including authenticated fuzzing with raw requests, auto-calibration, and result analysis

Catch up on uncommitted work — read all uncommitted git changes (staged, unstaged, untracked) into the conversation to establish context. Use when resuming work or when the user runs /catchup.

Bring a JavaScript repo up to the standard Code Quality Playbook baseline (Prettier, ESLint ratchet, husky + lint-staged, CI lint gate, optional JSDoc typecheck) idempotently — present pieces are skipped, missing ones are added — then optionally migrate the repo from npm to pnpm (lockfile import, packageManager pin, build-script allow-list, Firebase functions-framework fix, npm cleanup). Manual, global. Use only when the user runs /code-quality.

Create git commits for staged and unstaged changes using the emoji commit-message format, with optional push. Use when the user wants to commit or check in changes, or runs /commit.

Execute one complete unit of work end-to-end through tight feedback loops — understand it, plan it, build it, validate it, commit it — never advancing a phase until its feedback signal is green. Manual and global: runs only when the user invokes /do-work or another skill explicitly hands it a scoped unit of work. Use to "do the work", "execute this task/plan/issue", "build and validate and commit", or as the execution engine other planning skills delegate to.

Set up Playwright E2E testing framework for any project — auto-detects stack, auth, routes, and selectors

Generate a structured development epic through an interactive, phase-by-phase approval workflow that breaks a large piece of work into tasks. Use to create an epic, break down a large feature, or when the user runs /epic.

Set up git guardrails — a Claude Code PreToolUse hook that intercepts and blocks dangerous git commands such as force-push and hard reset before they run. Use to install git safety hooks or when the user runs /git-guardrails.

Interview the user relentlessly about a plan or design until reaching shared understanding, resolving each branch of the decision tree. Use when user wants to stress-test a plan, get grilled on their design, or mentions "grill me".

Work through open GitHub issues one at a time, each executed as a complete unit of work through a feedback loop (TDD then validate, never advancing on red) behind an enforced lint-staged and Prettier quality gate, then shipped as a feature branch and PR. Use to grind issues or when the user runs /grind.

Break a PRD into independently-grabbable GitHub issues using vertical slices (tracer bullets). Use to convert a PRD or spec into actionable issues or when the user runs /prd-to-issues.

Turn a PRD into a phased implementation plan using tracer-bullet vertical slices, saved as a Markdown file in ./plans/. Use to plan implementation from a PRD or when the user runs /prd-to-plan.

Rapidly file one or more well-structured, agent-friendly GitHub issues in the current repo. Use to quickly capture bugs or tasks as issues or when the user runs /qa.

Run accessibility and visual design review

Build features or fix bugs using red-green-refactor TDD with vertical slices, where tests verify behavior through public interfaces. Use for test-driven development or when the user runs /tdd.

Build and maintain a project's ubiquitous language — a CONTEXT.md glossary (plus ADRs and, for multi-repo projects, a context map). Interviews you to sharpen terminology, grounds terms in the code, and keeps the docs a living, drift-checked source of truth. Use to bootstrap or audit domain terminology, or when the user says "/terms".

Investigate a reported problem, find its root cause, and file a GitHub issue with a TDD fix plan — a mostly hands-off workflow. Use to triage a bug or diagnose root cause, or when the user runs /triage-issue.

Create a Product Requirements Document through an interactive interview, then file it as a GitHub issue. Use to write a PRD or spec or when the user runs /write-prd.