agentleFS
Wolfe-JamWolfe-Jam/faf-skills/skills/faf-context/SKILL.md

faf-context

Skill4 starsChanged 4 days agoMITView on GitHub

Runs `faf auto` to detect the stack, suggests the goal and six W''s from what the repo already says and asks you to confirm each one, scores the project.faf 0–100%, and `faf sync` writes the facts into CLAUDE.md. Use when someone wants Claude to know their project from the first message: "improve my CLAUDE.md", "set up project context", "Claude keeps forgetting my stack". Takes a CLAUDE.md from GOOD (it exists) to BETTER (backed by a scored project.faf). The quickest path; faf-go fills the last gaps, faf-expert goes deep.

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 1

    1. Wolfe-Jam/rust-faf-mcp★ 4same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository7 other files it gives its agents.

The file

Files
SKILL.md
<!-- faf:start -->
v: 1
needs:
  - "faf-cli on Node 22+ (not installed: run `npx faf-cli <command>` in place of `faf <command>`)"
reads: [project.faf, "the repo's manifests and README"]
writes:
  - "project.faf (faf auto)"
  - "CLAUDE.md (faf sync, faf's block only)"
network: ["registry.npmjs.org (read, only when npx fetches faf-cli)"]
publishes: false
done_when: "project.faf is scored, and CLAUDE.md is written by faf sync"
project: project.faf
<!-- faf:end -->

FAF Context — Give the AI What It Needs

AI writes its best code when it has your project's context. This skill helps you hand it over — fast, and to 100%.

.faf is an IANA-registered context format (application/vnd.faf+yaml) — a typed, portable file you own, readable by any AI (no bespoke manifest, no vendor lock-in). The whole point is one number: AI-readiness, 0–100%. At 100% ✪ the AI starts every session already knowing your project — no re-explaining, no guessing. This skill is the builder's path to that number: minimum typing, maximum context.

For a .faf generated from the repo, with the goal and six W's suggested for you to confirm, use faf-wizard. To master the format, use faf-expert. This skill is the quickstart in between.

How it works: app-type → AI fills the max → you answer the gaps

Every .faf is scored against the same 33 slots. faf-cli fills the 21 base slots and marks the 12 enterprise slots slotignored unless your app type uses them; slotignored slots never count against you. 100% ✪ = every active slot filled. It works in three steps, and only the last one needs you:

  1. Your app type sets which slots are required. A CLI needs different slots than a full-stack web app — faf-cli right-sizes the set and slotignoreds the rest (never counted against you).
  2. The AI fills what it can detect. faf auto detects your stack and language. The goal and six W's it only suggests, from your README and package.json, and a sharp goal sentence seeds suggestions for who/what/where.
  3. You confirm the rest. faf go shows each suggestion with where it came from; you accept it, edit it, or skip. Anything with no suggestion, you answer. → 100% ✪.

So your job isn't "fill 33 boxes." It's: write one good goal, confirm the suggestions, and answer the few the repo can't.

You rarely type all 6 Ws — here's why

The 6 Ws are "the underivable half": they're yours, and nothing is written into them until you confirm it. But you rarely type all six from scratch:

  • who / what / where → suggested from your goal sentence (the facts your goal literally states)
  • why / how → suggested from your README when it has a ## Why or ## How it works section
  • when → usually yours to type: the stage it's at, and since when

So the fast path is: write one sharp goal, confirm the suggestions, type what's left. → 100%.

"Sometimes the goal alone is enough." A great goal sentence and a README with a Why section can suggest most of the six, leaving you one or two short answers. You still confirm each one.

The fastest path to 100% ✪

faf auto      # 1. Detects your stack; says how many suggestions wait for the goal and 6 Ws
faf score     # 2. See the number + exactly which slots are still empty
faf go        # 3. Each suggestion in gray, with its source: Enter confirms, Tab edits, skip leaves it
faf score     # 4. 100% ✪
faf sync      # 5. Push context into CLAUDE.md (optional; AGENTS.md: faf export --agents)

Most projects are one good goal sentence, a few confirmations and an answer or two away from Trophy.

Write the ONE goal sentence (this does the heavy lifting)

The goal is the generative input: it seeds suggestions for who/what/where, which you confirm in faf go. Make it a real, specific sentence (it's also your use-case):

  • ✓ "A CLI that scores any repo's AI-readiness and syncs context to Claude, Cursor, and Gemini — for solo developers." → suggests what (a CLI that scores AI-readiness), where (Claude, Cursor, Gemini), who (solo developers). You confirm those and add why, when and how.
  • ✗ "A tool to improve development." → generic; seeds nothing. (Generic phrases are ignored on purpose — empty beats wrong.)

The 6 Ws — one line each, every word earns its place

Each W is one line of fact, not a paragraph: short when the fact is short, a full line when it needs one.

WHoldsExample
Whothe people who use it, not the authorsolo developers
Whatwhat they make or do with itan AI-readiness score for any repo
Whythe problem it removesAI forgets the project every session
Wherewhere it runs or shipsnpm, Homebrew
Whenthe stage it's at, and since when (not a version)production since September 2025
Howhow it works: the mechanism, not an install command or a stack listdetects the stack from repo files; scores with a WASM kernel

Slots that don't apply → slotignored

A CLI has no frontend; an API has no UI library. Mark those slotignored and they drop out of the denominator — you're scored only on slots that matter for your app type. 100% means "everything that applies is filled," not "every box checked." (faf auto and faf go handle most of this for you.)

The honesty rule (why this works)

The AI only suggests facts your goal/README literally state, never invents, never uses templates, and writes none of the goal or six W's until you confirm it. What it can't source, it leaves empty for you. Empty beats wrong. That's why the resulting context is trustworthy: every slot is either detected, confirmed by you, or honestly blank.

When you're done here

  • Want a .faf generated from the repo, with the goal and six W's suggested for you to confirm? → faf-wizard
  • Want to master the format (scoring internals, MCP config, sync)? → faf-expert
  • Driving a repo all the way with an agent? → faf go / faf loop

The goal: the AI is only as good as the context you give it. Answer the few things only you know — the gaps it couldn't fill itself — and it's optimized to help you at 100% ✪.

MIT · part of the FAF skill family (faf-context · faf-wizard · faf-expert)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.