faf-expert
Skill44 starsChanged 10 months agoMITView on GitHub
Expert in the .faf format (Persistent Project Context for AI) and the faf-cli toolkit. Use when working with .faf files, scoring AI-readiness, syncing project context to CLAUDE.md, or discussing the Foundational Context Layer (FCL) doctrine. IANA-registered (application/vnd.faf+yaml). Anthropic-listed as "Persistent Project Context Server" (PR #2759, merged Oct 17 2025). Current CLI: v6.6.0 — The Trophy Edition.
- Installs packages
2 other projects use a version of this.
Adopted by
Most recently taken up by
- Wolfe-Jam/faf-skillstheir own version, first seen Nov 13, 2026
- Wolfe-Jam/grok-faf-mcptheir own version, first seen Jan 26, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- Wolfe-Jam/grok-faf-mcp★ 19same name
- Wolfe-Jam/faf-skills★ 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 repository3 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- .github/copilot-instructions.mdCopilot instructions
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 55
Run faf --help for the full 26-command set.
- Distinctive: line 100
Read the rationale; refute by inspection.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
FAF Expert — Persistent Project Context Specialist
What is .faf?
The .faf file is Persistent Project Context for AI — a single source of truth that survives across sessions, tools, and AI assistants.
- IANA-registered media type:
application/vnd.faf+yaml(registered Oct 30, 2025) - Anthropic-listed: modelcontextprotocol/servers PR #2759 — claude-faf-mcp shipped as "Persistent Project Context Server" (merged Oct 17, 2025)
- Git-native:
project.fafversions with your code; every clone gets full AI context - The CLI behind the FCL — the Foundational Context Layer for AI
The credo: FAF defines. AGENTS.md instructs. AI codes.
When to activate this skill
- Creating, editing, or scoring
.faffiles - Syncing
.faf→CLAUDE.md, or exporting it toAGENTS.md,.cursorrules,GEMINI.md - Diagnosing sub-Trophy scores and reaching 🏆 100%
- Explaining the FCL, tier system, or app-type ladder
- Connecting
.fafwork to the broader receipts: IANA, Anthropic PR #2759, MCP Registry
Install the CLI
bunx faf # Bun — zero install, fastest
npx faf # npm — works everywhere
npm install -g faf-cli # global install
brew install faf-cli # Homebrewfaf is the verb across every surface: bunx faf, npx faf, /faf (Claude Code plugin), faf (global). Bare faf shows your project's score; faf auto detects and fills.
The workflow
faf init → faf auto → faf go → 🏆 Trophy 100%| Command | What it does |
faf init | Create a project.faf scaffold for your project |
faf auto | Auto-detect stack, fill what's detectable, score |
faf go | Guided interview to fill the remaining slots → Trophy |
faf score | Check current AI-readiness (0–100%) |
faf sync | Write CLAUDE.md from .faf (one way) |
faf compile | .faf → .fafb binary (sealed, portable, deterministic) |
faf check | Validate .faf structure |
faf info | Version + system info |
Run faf --help for the full 26-command set.
Scoring — Trophy 100% all-or-nothing
From v6.6.0 onward, faf-cli recommends only 🏆 Trophy.
| Tier | Score | Status |
| 🏆 Trophy | 100% | AI never has to guess |
| ★ Gold | 99%+ | 1 slot from Trophy |
| ◆ Silver | 95%+ | Close — keep going |
| ◇ Bronze | 85%+ | Interim — keep going |
| ● Green | 70%+ | Interim — keep going |
| ● Yellow | 55%+ | AI flipping coins |
| ○ Red | <55% | AI working blind |
| ♡ White | 0% | No context at all |
Why all-or-nothing? Architectural, not aspirational:
Layer 4 AI tooling ← AI optimised by complete FCL
Layer 3 Agents ← can act because FCL is complete
Layer 2 MD instructions ← can be re-authored correctly
Layer 1 .faf (FCL) ← 🏆 Trophy = complete = foundation100% on Layer 1 makes Layers 2–4 work. Sub-100% degrades every layer above — instructions miss context, agents guess, AI optimisation can't happen.
🏆 is the only emoji. Sub-Trophy tiers use geometric Unicode (★ ◆ ◇ ● ○ ♡). The retired medal ladder (🥇🥈🥉🟢🟡🔴🤍) is history.
The app-types
The app-type answers one question: what does this repo ship? It sets which of the 33 slots count; the rest are slotignored. App-types are types of projects, not types of file (a CLAUDE.md or a SKILL.md sits inside a project of some type).
The full table, with each type's slots, is in docs/SCORING.md. Write the type's own name (cli, not cli-tool); a name faf doesn't know is treated as library.
About (not an app_type): declare about.represents: owner/repo. Inherits about.source_score. Do not set app_type: about.
The CLI auto-detects via # found: rationale next to type::
project:
type: cli # found: package.json binRead the rationale; refute by inspection. Glass Hood doctrine — the CLI shows its work.
.faf file structure (current schema)
faf_version: 3.0
app_type: cli
project:
name: my-project
goal: One-line elevator pitch
main_language: TypeScript
stack:
framework: React
backend: Node.js
db: PostgreSQL
api: REST
css: Tailwind
state: Zustand
pkg_manager: bun
hosting: Vercel
build: vite
cicd: GitHub Actions
runtime: Node 20+
connection: prisma
human_context:
who: target users
what: one-line description
why: the problem this solves
where: where it ships
when: production since YYYY-MM
how: how it works — the mechanism, and the choice that shapes the codev6.6.0 uses Mk4 canonical slot names in user-facing docs: db (was database), framework (was frontend), css (was css_framework), state (was state_management), api (was api_type), pkg_manager (was package_manager). 36k+ existing .faf files continue scoring correctly via read-time aliasing.
The sync flow — one way
.faf is the canonical Foundational Context Layer (FCL). CLAUDE.md / AGENTS.md / .cursorrules / GEMINI.md are downstream prose renders that READ .faf to save AI time. They never write back automatically.
.faf ──── 8ms ───→ CLAUDE.md (faf sync)
AGENTS.md (faf export)
.cursorrules
GEMINI.mdfaf sync --direction pull (CLAUDE.md → .faf backfill of name, goal and language) is Trophy-gated — blocked below 100% to prevent overwriting canonical slots with prose drift.
Receipts ladder
- IANA-registered:
application/vnd.faf+yamlat iana.org/assignments/media-types/application/vnd.faf+yaml — registered 2025-10-30 - Anthropic-listed: PR #2759 merged Oct 17, 2025 — claude-faf-mcp added as "Persistent Project Context Server"
- MCP Registry: FAF Ecosystem #2759 (Five Fingers: claude / faf / grok / gemini / WJTTC)
- One dev, <12 months — IANA-registered, Anthropic-listed, xAI ZEPH commission, 100+ FAF-disciplined repos, the format used to build itself
Where to go next
| You want to... | Go here |
| Install + start using the CLI | github.com/Wolfe-Jam/faf-cli |
| Read the v6.6 release | CHANGELOG — The Trophy Edition |
| Understand the FCL / Trophy doctrine | docs/SCORING.md |
| Set up sync | docs/SYNC.md |
| Visit the format home | faf.one |
| Pin context to AI tools | Run faf export — authors AGENTS.md, .cursorrules, GEMINI.md from your .faf |
FAF defines. AGENTS.md instructs. AI codes. v6.6.0 — The Trophy Edition. How no-score became a score.
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
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.

