Best use case
fast is best used when you need a repeatable AI agent workflow instead of a one-off prompt.
Zero-ceremony inline execution for 3 or fewer file edits.
Teams using fast should expect a more consistent output, faster repeated execution, less prompt rewriting.
When to use this skill
- You want a reusable workflow that can be run more than once with consistent structure.
When not to use this skill
- You only need a quick one-off answer and do not need a reusable workflow.
- You cannot install or maintain the underlying files, dependencies, or repository context.
Installation
Claude Code / Cursor / Codex
Manual Installation
- Download SKILL.md from GitHub
- Place it in
.claude/skills/fast/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How fast Compares
| Feature / Agent | fast | Standard Approach |
|---|---|---|
| Platform Support | Not specified | Limited / Varies |
| Context Awareness | High | Baseline |
| Installation Complexity | Unknown | N/A |
Frequently Asked Questions
What does this skill do?
Zero-ceremony inline execution for 3 or fewer file edits.
Where can I find the source code?
You can find the source code on GitHub using the link provided at the top of the page.
SKILL.md Source
# /fast - Zero-Ceremony Execution The Fast tier sits at the bottom of the five-tier task hierarchy (Fast > Quick > Simple > Medium > Complex). It exists because the full ceremony of plan files, agent routing, and quality gates is wasteful for a typo fix. Execute inline without plans or subagents, commit the result, and log it. --- ## Instructions ### Phase 1: UNDERSTAND **Goal**: Confirm the task is Fast-eligible and know exactly what to change. **Step 1: Read CLAUDE.md** Read and follow the repository CLAUDE.md before any other action, because repository-specific constraints may affect how the edit should be made. **Step 2: Parse the request** Identify from the user's request: - Which file(s) need editing - What specific change is needed - Whether this is clearly a 1-3 edit task If `--dry-run` was passed, show what would change without editing and stop. **Step 3: Scope check** Ask these questions silently (do not display to user): | Question | If Yes | |----------|--------| | Does this need reading docs, investigating behavior, or understanding unfamiliar code? | Redirect to `/quick --research` because investigation means uncertainty, and uncertainty means this is not a Fast task | | Does this touch more than 3 files? | Redirect to `/quick` because uncapped "fast" tasks silently grow into untracked large changes | | Does this add imports from new packages, install libraries, or modify dependency files (go.mod, package.json, requirements.txt)? | Redirect to `/quick` because new dependencies carry risk that needs proper tracking | | Is the change ambiguous or underspecified? | Ask user one clarifying question. If still ambiguous after one round, redirect to `/quick --discuss` | If redirecting, say: ``` This task exceeds /fast scope ([reason]). Redirecting to /quick. ``` Then invoke the quick skill with the original request. **Step 4: Locate target files** Read the file(s) that need editing. Confirm the exact lines to change. **GATE**: Task is confirmed Fast-eligible (1-3 edits, no research, no new deps). Target files identified and read. ### Phase 2: DO **Goal**: Make the edits inline without spawning subagents, because subagents add latency and context setup cost that dwarfs the actual work at this tier. Do not create `task_plan.md`, because the overhead of planning exceeds the task itself for a 1-3 edit change. **Step 1: Execute edits** Make the changes using the Edit tool. Track the number of files edited after each operation, because the 3-edit scope gate depends on an accurate count. **Step 2: Mid-execution scope check** After each edit, check: have we hit 3 edits? If the task needs MORE edits to complete, stop immediately — do not rationalize "just one more edit" because the 3-edit gate exists specifically to prevent silent scope creep: ``` Scope exceeded during execution (3+ edits needed). Preserving work done. Redirecting remainder to /quick. ``` Hand off to `/quick` with context about what was already done. Do not start additional "while I'm here" fixes, because scope creep in fast mode produces untracked large changes with no plan or review. **GATE**: All edits complete. Edit count is 1-3. ### Phase 3: COMMIT **Goal**: Commit the changes with a clean message, because uncommitted fast edits are invisible to the audit trail and defeat the traceability that justifies using the system at all. Even one-line changes get commits because one-line changes cause one-line bugs that are invisible without commit history. If `--no-commit` was passed, skip this phase (for when the user wants to batch changes). **Step 1: Check branch** If on main/master, create a feature branch first because even fast tasks get proper branches — never commit directly to main: ```bash git checkout -b fast/<brief-description> ``` **Step 2: Stage and commit** ```bash git add <specific-files> git commit -m "$(cat <<'EOF' <type>: <description> EOF )" ``` Use conventional commit format. The type is usually `fix:`, `chore:`, or `refactor:` for fast tasks. **GATE**: Commit succeeded. Verify with `git log -1 --oneline`. ### Phase 4: LOG **Goal**: Record the task for auditability, because without logging fast tasks are invisible and the system loses its audit trail. **Step 1: Append to STATE.md** If STATE.md exists in the repo root, append to the quick tasks table. If it does not exist, create it. Format: ```markdown ## Quick Tasks | Date | ID | Description | Commit | Tier | |------|----|-------------|--------|------| | YYYY-MM-DD | - | <description> | <short-hash> | fast | ``` Fast tasks do not get task IDs (that is a Quick-tier feature). Use `-` for the ID column. **Step 2: Display summary** ``` =================================================================== FAST: <description> =================================================================== Files edited: <N> Commit: <hash> on <branch> Logged: STATE.md =================================================================== ``` --- ## Error Handling ### Error: Scope Exceeded Mid-Execution **Cause**: Task turned out to need more than 3 edits **Solution**: Stop, preserve work, redirect to `/quick` with context about completed edits. Do not undo work already done. ### Error: Ambiguous Request **Cause**: Cannot determine exact files or changes from the request **Solution**: Ask user one clarifying question. If still ambiguous after one round, redirect to `/quick --discuss`. ### Error: On Main Branch **Cause**: Currently on main/master **Solution**: Create `fast/<description>` branch before editing. Never commit directly to main.
Related Skills
x-api
Post tweets, build threads, upload media via the X API.
worktree-agent
Mandatory rules for agents in git worktree isolation.
workflow
Structured multi-phase workflows: review, debug, refactor, deploy, create, research, and more.
workflow-help
Interactive guide to workflow system: agents, skills, routing, execution patterns.
wordpress-uploader
WordPress REST API integration for posts and media uploads.
wordpress-live-validation
Validate published WordPress posts in browser via Playwright.
with-anti-rationalization
Anti-rationalization enforcement for maximum-rigor task execution.
voice-writer
Unified voice content generation pipeline with mandatory validation and joy-check. 8-phase pipeline: LOAD, GROUND, GENERATE, VALIDATE, REFINE, JOY-CHECK, OUTPUT, CLEANUP. Use when writing articles, blog posts, or any content that uses a voice profile. Use for "write article", "blog post", "write in voice", "generate content", "draft article", "write about".
voice-validator
Critique-and-rewrite loop for voice fidelity validation.
vitest-runner
Run Vitest tests and parse results into actionable output.
video-editing
Video editing pipeline: cut footage, assemble clips via FFmpeg and Remotion.
verification-before-completion
Defense-in-depth verification before declaring any task complete.