claude-forge

Use when you need to commit and push changes — stages files, commits with co-author tags, pulls with rebase, resolves conflicts, and pushes. Also supports "pr" mode for creating PRs and "clean" mode for pruning merged branches. Also use for "commit", "push", "sync", or "create PR".

Use when the user wants short-and-sweet answers to one or more questions. Restates each question as a bold heading, then answers in 2–5 lines per question. Bullets and small tables allowed when they help. Also use for "quick question", "qq", "answer briefly", or any "/my-qq <question>" invocation.

Use when starting implementation on any project — reads planning docs, dispatches sub-agents, and creates or reiterates docs/acceptance-criteria.md with measurable criteria per feature. Must run before writing implementation code. Also use for "create acceptance criteria", "what are the requirements", or "define done".

Use when starting a new feature or project from a raw idea — asks targeted questions and produces a structured PRD at docs/prd-<slug>.md. Sits upstream of /my-prompt. Also use for "write a PRD", "product requirements", or "what should we build".

Use when code needs rigorous review before shipping — spawns skeptic/believer/referee agents in iterative rounds until findings converge to nitpicks. Use after implementing features, before merging PRs, or when the user asks for a "thorough review", "adversarial review", or "tear this apart".

Use when facing a complex task that benefits from multiple specialized agents working in parallel — spawns, delegates, tracks status, and collects results. Also use when the user says "use agents", "multi-agent", or "parallelize this".

Use when you want to find and fix gaps in the codebase — dispatches all agent personalities to audit code against CLAUDE.md guardrails and acceptance criteria, documents gaps with fix suggestions, then implements fixes. Also use for "audit the code", "find gaps", or "check against criteria".

Use when you need to fetch and extract a web article's content as clean markdown — handles blogs, docs, PDFs, arxiv. Also use for "fetch this article", "read this URL", or "extract content from".

Use when a PR needs thorough review — dispatches 6 specialized agents (silent failures, type design, simplification, test gaps, comment analysis, general review) in parallel. Also use for "review this PR", "review PR

Use when ending a session or at a natural checkpoint — saves session summary, updates state.md, copies transcript, extracts learnings, and triggers consolidation when 5+ sessions accumulate. Also use for "save", "save session", or "checkpoint".

Use when you need to check your Claude Code setup — quick lookup for installed skills, hooks, agents, session info, settings, and plugins. Also use for "show info", "what skills do I have", "list hooks", or "my setup".

Use when you need a security audit — runs SAST, dependency audit, secret detection, and STRIDE threat modeling on the current project. Also use for "security scan", "check for vulnerabilities", "audit security", or "find secrets".

Use when the user wants to restrict edits to a specific directory — blocks Edit/Write operations outside the allowed path. Invoke as "/my-freeze src/components" to lock edits to that directory only. Use "/my-freeze off" to deactivate.

Use when starting an implementation session on any project — creates or updates features.json, runs init checks, verifies environment, and sets up clean state for the worker session. Also use for "init session", "set up for implementation", or "prepare to code".

Use when you want to find bugs before shipping — runs 3 adversarial agents (Hunter finds bugs, Skeptic disproves false positives, Referee arbitrates) against your codebase or specific files. Also use when the user says "find bugs", "hunt for issues", or "check for problems".

Use when existing code lacks tests — analyzes functions, branches, and edge cases then generates comprehensive test suites. Also use for "write tests", "add test coverage", "generate tests for this", or "backfill tests".

Use when you need to QA a running web app — launches a browser, explores pages, finds bugs, and generates regression tests for issues found. Also use for "QA the app", "test the UI", "browser test", or "find bugs in the frontend".

Use when you want to save the session AND push changes in one command — runs /my-save then /my-git-sync sequentially. Also use for "save and push", "save and sync", or "wrap up and push".

Use when you want to analyze Claude session history and surface automation opportunities — finds repeated patterns, skill candidates, agent workflows, config improvements, and coding style fixes. Also use for "self improve", "analyze sessions", "what should I automate", or "find patterns".

Use when you want to improve a skill's quality automatically — runs iterative self-improvement loop with binary checklist scoring. Adapts Karpathy's autoresearch method. Also use for "improve this skill", "optimize skill", or "autoresearch".

Use when a PR needs monitoring — watches CI status, retries flaky tests, resolves merge conflicts, and optionally auto-merges when ready. Use when the user says "watch this PR", "babysit the PR", or "let me know when CI passes".

Use when you need architecture diagrams — generates C4 diagrams (Context, Container, Component, Code) from codebase analysis as Mermaid. Also use when the user says "draw the architecture", "system diagram", or "C4 diagram".

Use when about to do dangerous operations (deploys, database migrations, force pushes, production changes) — activates on-demand hook that blocks destructive commands. Also use when the user says "be careful", "careful mode", or wants extra safety. Use /my-careful again to deactivate.

Use when starting a session and need context on recent work — reads last N session summaries and presents a formatted overview of what was done, decisions made, and current status. Also use for "what did I do last time", "catch me up", or "recent sessions".

Use when you need to fetch a GitHub repo's metadata, README, and file tree — wraps gh CLI. Returns structured data. Also use for "check this repo", "what is this repo", or "fetch repo info".

Use when you need a quick overview of recent work across sessions — reads session summaries and presents colored output of what was done, key decisions, and current status. Also use for "what's been happening", "get me up to speed", or "session history".

Use when you need to manage hooks without editing settings.json — creates, lists, updates, or removes Claude Code hooks via natural language. Also use for "add a hook", "list hooks", "create hook for", or "remove hook".

Use when there's a production incident or outage — structured triage, investigation, fix, verification, and postmortem. Also use for "site is down", "production issue", "incident", or "outage".

Use when building LLM-powered features — provides guided patterns for RAG, embeddings, prompt engineering, evaluation pipelines, and agent architectures. Also use for "build with LLM", "add AI feature", "RAG pipeline", or "prompt engineering".

Use when you have multiple tasks to execute sequentially — generates disciplined prompts via /my-prompt, then executes each task with plan→criteria→build→review→docs→checkpoint cycle. Also use for "run these tasks", "execute the list", or "loop through tasks".

Use before starting any significant feature or project — adversarial problem reframing that challenges your assumptions, explores alternative framings, and finds the simplest path before you commit to a design. Also use for "office hours", "challenge this idea", "am I solving the right problem", or "reframe this".

Use when creating new Claude Code extensions — scaffolds, validates, and tests skills, agents, hooks, and commands. Also use for "create a skill", "new hook", "build agent", or "scaffold plugin".

Use when you need to see what prompts were used recently — shows last N user prompts from current session or all sessions. Also use for "show my prompts", "what did I ask", or "prompt history".

Use when about to start a significant coding task — transforms rough thoughts into a disciplined prompt with acceptance criteria, anti-sycophancy rules, and complexity budget. Also use for "write a prompt", "create prompt for", or "discipline this task".

Use when planning docs need review before implementation — dispatches all agent personalities (skeptic, believer, architect, security) against docs in parallel, aggregates findings by priority. Also use for "review the docs", "check planning docs", or "doc review".

Use when debugging production issues, diagnosing service health, or investigating symptoms like "the app is slow", "API returns 500", "Docker container keeps restarting", "database connection refused". Structured symptom → investigation → diagnosis → report.

Use when a project needs e2e smoke tests — generates Playwright test files that live in the project repo and run inside Docker. Also use for "write smoke tests", "add e2e tests", "generate tests for the UI", or "test the app".

Use when you want to change the statusline display — switches between mini (compact), max (full), or none (disabled) modes. Also use for "change statusline", "compact status", or "hide statusline".

Use when testing local web apps with Playwright — browser automation, interaction testing, and screenshot verification. Also use for "test the app", "playwright test", or "browser automation".