ef-feature-docs
Stage 2 - author a feature's documentation folder from closed grill decisions: grill-decisions to RFC to PRD to INDEX, with mandatory fields gated by asking…
Skill by micaelcf
Chalk Labs
Need help deploying ef-feature-docs?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/micaelcf-ef-feature-docs
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: writing
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md and code. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/ef-feature-docs
What's inside
- Skill name
ef-feature-docs- Description
- Stage 2 - author a feature's documentation folder from closed grill decisions: grill-decisions to RFC to PRD to INDEX, with mandatory fields gated by asking rather than invented, stable requirement identifiers, and an amendment protocol that keeps the documents alive through implementation. Use after a product-altitude grilling has closed the decisions, or when the user says write the RFC, write the PRD, author the feature docs.
- License
- MIT
- Instructions
- 108 lines in SKILL.md · view on GitHub
Files (6)
- SKILL.md
- amendments.md · references/
- anti-patterns.md · references/
- index-template.md · references/
- prd-template.md · references/
- rfc-template.md · references/
SKILL.md (first lines)
# Feature docs — Stage 2 of 5 Turn closed decisions into the document set every downstream repository reads. You are **transcribing and structuring decisions already made**. You are not re-deciding them, not designing the implementation, and not filling gaps with plausible content. ## Order is the mechanism ``` grill-decisions (exists) → RFC → PRD → INDEX ``` Each document is written **from** the one before, never beside it: - **`grill-decisions`** is the rationale source of truth and outranks everything written from it. Stage 1 produced it; you do not modify it. - **The RFC** carries the decisions with their options, criteria, assumptions and risks. Its decision anchors `D1..Dn` map 1:1 to the grill's `Q1..Qn`. Where a grill question produced no anchor, say so rather than renumbering. - **The PRD** carries requirements with stable identifiers, acceptance criteria and success metrics. It is written from the RFC's decisions, not from the grill directly. - **`INDEX.md`** is the map: what this feature is, which document settles what, the reading order, the decision index, and the per-repo status. Written in parallel they drift, and the RFC's own precedence header becomes a lie on the day it is written. ## Gate before you write Never generate a document with an invented value in a mandatory field. Each template names its mandatory fields and the question to ask when one is missing. Collect what is missing with the `ask` tool — highest-consequence question first, one wave at a time, not a single long form. If the user provided rich context, use it; do not ask for what you already have. Anything only the user can settle that asking did not resolve goes to `## Open questions` with an owner and a deadline. **A placeholder that reads like content is the one failure this stage must not produce** — a downstream repo will build against it. ## The PRD is the identifier registry Stage 4 extracts checks by requirement identifier and Stage 5 audits by walking them. So: - every requirement gets a stable id (`RF-###` or the repo family's convention) - ids are **never renumbered**; a superseded requirement is struck through and kept - a requirement with no id cannot be traced, proven or audited, and will be silently dropped ## These documents are alive Do not author a "final" RFC. Stage 3 amends it when an anchor comes back false, Stage 4 when the shipped shape differs from the design, Stage 5 when something did not ship. Its header is the only home for cross-repo delivery status. Read `references/amendments.md` before writing the header, so the document has somewhere for that to land. ## Language Author in **English** unless the user, or the repository's `ef-harness` block, names another language. Product copy quoted inside the documents keeps its own locale. Technical terms stay in English regardless. ## Scope This skill authors **once**. It is a post-grill step, not a regenerator: after the first authoring, these documents change only through amendments. That is why there is no merge behaviour to define. ## Scale Read the **scale** declared in the grill-decisions header. At `standard` and `multi`, author the full set: RFC, PRD, INDEX. At `single` — one repository, no one-way door — collapse the RFC and the PRD into **one document**, `DECISIONS_<feature>.md`, and skip INDEX unless the folder holds more than two files. Keep: - the decisions with their anchors, evidence grades, rejected options and rationale - the assumptions table, with evidence grades and invalidation triggers - the requirements, with stable identifiers and acceptance criteria - non-goals, risks and open questions Drop the cross-repo apparatus: target-repo status rows, per-repo contributors, and the handoff document. There is one repo, so the code is the contract and the audit reads it directly. Two documents for a four-requirement feature cost tokens at every downstream stage and hide nothing extra. The amendment protocol works unchanged on one file. ## References — load at the step that needs them | File | Load when | |---|---| | `references/rfc-template.md` | writing the RFC | | `references/prd-template.md` | writing the PRD | | `references/index-template.md` | writing INDEX.md | | `references/amendments.md` | writing the RFC header, or amending later | | `references/anti-patterns.md` | before finalising either document | Do not load them during the first pass over the grill decisions. ## Output State where each file was written, which mandatory fields you had to ask about, and what remains in `## Open questions` with an owner. Then name the next stage: `/ef:spec-drift` or `/ef:perspective-discovery`, per target repo.
From micaelcf/ef-harness, MIT.
More from micaelcf/ef-harness
- 0.610score
ef-readiness
micaelcf
Make a repository's agent rulebook (AGENTS.md / CLAUDE.md) verifiable: vendor a deterministic checker, audit every rule and citation against the code…
SkillAFreshness: A★ 4Claude Code +9 - 0.610score
ef-implement
micaelcf
Stage 4 - turn closed decisions and a written contract into checks with named proofs, then orchestrate the build in parallel slices and hand a whole feature…
SkillAFreshness: A★ 4Claude Code +9 - 0.610score
ef-grilling
micaelcf
Relentless one-question-at-a-time interview that converts an under-specified plan into closed decisions, each recorded with its evidence grade, preserved…
SkillAFreshness: A★ 4Claude Code +9 - 0.610score
ef-adversarial-audit
micaelcf
Stage 5 - adversarially audit delivered work against its requirements assuming a green test suite proves nothing, then verify that claimed fixes are real…
SkillAFreshness: A★ 4Claude Code +9 - 0.609score
ef-prototype
micaelcf
Discover a brand-new UI before implementing it: generate 4-5 radically different variants of one screen in a single shareable HTML file, switchable from a…
SkillAFreshness: A★ 4Claude Code +9 - 0.606score
ef-shared
micaelcf
Shared vocabulary for every ef-* stage: the A–D evidence grades and the five critical lenses (assumption-surfacing, pre-mortem, red-team, evidence-audit…
SkillAFreshness: A★ 4Claude Code +9 - 0.605score
ef-perspective-discovery
micaelcf
Discover a feature from the current repository's perspective, after it was defined upstream at product altitude: verify the docs' code anchors, fan out…
SkillAFreshness: A★ 4Claude Code +9
Signals
- Stars
- 4
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Sep 28, 2026 (1 month ago)
- Last release
- Oct 21, 2026 (15 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/micaelcf/ef-harness
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from micaelcf
- 0.610score
ef-readiness
micaelcf
Make a repository's agent rulebook (AGENTS.md / CLAUDE.md) verifiable: vendor a deterministic checker, audit every rule and citation against the code…
SkillAFreshness: A★ 4Claude Code +9 - 0.610score
ef-implement
micaelcf
Stage 4 - turn closed decisions and a written contract into checks with named proofs, then orchestrate the build in parallel slices and hand a whole feature…
SkillAFreshness: A★ 4Claude Code +9 - 0.610score
ef-grilling
micaelcf
Relentless one-question-at-a-time interview that converts an under-specified plan into closed decisions, each recorded with its evidence grade, preserved…
SkillAFreshness: A★ 4Claude Code +9 - 0.610score
ef-adversarial-audit
micaelcf
Stage 5 - adversarially audit delivered work against its requirements assuming a green test suite proves nothing, then verify that claimed fixes are real…
SkillAFreshness: A★ 4Claude Code +9 - 0.609score
ef-prototype
micaelcf
Discover a brand-new UI before implementing it: generate 4-5 radically different variants of one screen in a single shareable HTML file, switchable from a…
SkillAFreshness: A★ 4Claude Code +9 - 0.606score
ef-shared
micaelcf
Shared vocabulary for every ef-* stage: the A–D evidence grades and the five critical lenses (assumption-surfacing, pre-mortem, red-team, evidence-audit…
SkillAFreshness: A★ 4Claude Code +9
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get ef-feature-docs running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
ef-feature-docs skill. Agent Mart. https://agentmart.dev/skills/micaelcf-ef-feature-docs (last updated 2026-10-24)