feature-spec
Skill4 starsChanged 3 months agoMITView on GitHub
Write the executable feature specification — the WHAT and WHY artifact that agents and reviewers treat as source of truth. Owns both /specify and /clarify modes. Load when the user asks to write a feature spec, write a specification, write an executable spec, define functional requirements, capture acceptance criteria as Given/When/Then, or when the spec-driven-development orchestrator routes here. Also triggers on "feature spec", "executable spec", "/specify", "/clarify", "write the spec for this feature", "specification for", "spec-driven", "machine-readable spec". Output: docs/specs/YYYY-MM-DD-<slug>-feature-spec.md. Hard gate: cannot Approve while [NEEDS CLARIFICATION] markers remain.
21 other projects use a version of this.
Adopted by
Most recently taken up by
- insight68/Skillstheir own version, first seen Nov 8, 2026
- CesarRivasP/feature-spectheir own version, first seen Oct 15, 2026
- PiloTracer/pilo.ai.logicbisontheir own version, first seen Oct 11, 2026
- akachandan1/proofstacktheir own version, first seen Oct 11, 2026
- edu6565/claude-skillstheir own version, first seen Oct 11, 2026
- Simone-Tarantino/godot-superpowerstheir own version, first seen Oct 11, 2026
Used with
Projects with it also use changelog.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
HashiCorpSkillchangelog
Add a `.changelog/<PR_NUMBER>.txt` entry for a GitHub Pull Request, commit, and push.
Also in 4 projects that use this
Other versions21 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 21
- WikiEducationFoundation/WikiEduDashboard★ 431same name
- PackmindHub/packmind★ 321same name
- https-deeplearning-ai/sc-spec-driven-development-files★ 310same name
- natnew/awesome-ai-scientists★ 20same name
- yasunori0418/skills★ 9same name
- insight68/Skills★ 5same name
- carreiras/claude-skills★ 4same name
- devexpert-io/harness-starter★ 4same name
- gonimar/claude-web-studio★ 4same name
- Simone-Tarantino/godot-superpowers★ 2same name
11 more not listed
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 repository125 other files it gives its agents.
- AGENTS.mdAGENTS.md
- .cursor/rules/agent-loom-routing.mdcCursor rule
- .cursor/rules/agent-loom-skills-index.mdcCursor rule
- adversarial-hatSkill
- agent-builderSkill
- agent-launcherSkill
- agent-loom-syncSkill
- agent-observabilitySkill
- agent-run-retroSkill
- agent-system-architectureSkill
- api-and-interface-designSkill
- api-deprecation-and-migrationSkill
- apply-paper-to-projectSkill
- app-security-hardeningSkill
- architectural-decision-logSkill
- assumption-mappingSkill
- brainstormingSkill
- browser-testing-with-devtoolsSkill
- business-modelingSkill
- ci-cd-and-automationSkill
- codebase-understandingSkill
- code-review-crspSkill
- code-simplificationSkill
- compress-skillSkill
- context-engineeringSkill
- create-agent-promptSkill
- cross-link-skillsSkill
- customer-discoverySkill
- debug-and-fixSkill
- deep-thinkingSkill
- dependency-mappingSkill
- deploy-anywhereSkill
- deprecate-skillSkill
- design-directionSkill
- design-reviewSkill
- design-systemSkill
- dynamic-routingSkill
- eval-judgeSkill
- eval-outputSkill
- eval-pipelineSkill
- eval-rubric-designSkill
- experimentationSkill
- experiment-backlogSkill
- experiment-readoutSkill
- experiment-runbookSkill
- experiment-specSkill
- fault-localizeSkill
- fermiSkill
- first-principlesSkill
- frontend-designSkill
- generate-changelogSkill
- git-workflow-and-versioningSkill
- gsap-animationSkill
- harness-engineeringSkill
- harness-evolutionSkill
- harness-generationSkill
- idea-evaluationSkill
- idea-generationSkill
- implementation-planSkill
- improve-skillsSkill
- The first 60 shown.
The file
Show its rules and text
We split this skill into the rules it gives an agent
16 rules. 0 are common, 5 are distinctive, and 11 are very specific to this project.
0
Common
Most of the 19 edited versions we read keep these.
5
Distinctive
This author's own choices. Worth a look.
11
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 28
Never include architecture, library choices, file paths, or implementation details — those belong in implementation-plan.
- Very specific: line 29
Never mark status Approved while [NEEDS CLARIFICATION: ...] markers remain.
- Very specific: line 30
Never write the spec without referencing the project constitution version (docs/constitution.md@<N>).
- Very specific: line 31
Never invent functional requirements — if the user has not stated something, mark [NEEDS CLARIFICATION].
- Distinctive: line 32
Never use vague language ("fast", "intuitive", "robust") — replace with measurable criteria or mark for clarification.
- Distinctive: line 47
Ask only what cannot be inferred:
- Very specific: line 51
If the request is too vague to draft FRs, mark them [NEEDS CLARIFICATION] and continue — don't loop in interview.
- Very specific: line 57
Before drafting, if any requirement is inferred (stack, auth model, deployment target), list up to 5 bullets under ## Assumptions I'm …
- Very specific: line 59
Read references/feature-spec-schema.md for the full template.
- Very specific: line 104
Read the named (or latest) docs/specs/*-feature-spec.md.
- Distinctive: line 110
Wait for user answer.
- Very specific: line 111
Update the relevant FR/NFR/AC.
- Very specific: line 112
Remove CL-N from Needs Clarification list.
- Distinctive: line 116
After every answered CL, record internally:
- Distinctive: line 122
When CL list is empty AND all resolutions ≥70%, ask:
- Very specific: line 125
Only set Approved after explicit user confirmation.
“Common” is measured: kept by at least 6 in 10 of the 19 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Feature Spec
You are a Specification Engineer. You write feature specifications precise enough for an AI coding agent to plan from without follow-up questions, and structured enough for spec-crosscheck to validate. WHAT and WHY only — never HOW.
Hard Rules
Never include architecture, library choices, file paths, or implementation details — those belong in implementation-plan.
Never mark status Approved while [NEEDS CLARIFICATION: ...] markers remain.
Never write the spec without referencing the project constitution version (docs/constitution.md@<N>). If no constitution exists, offer to invoke project-constitution first.
Never invent functional requirements — if the user has not stated something, mark [NEEDS CLARIFICATION].
Never use vague language ("fast", "intuitive", "robust") — replace with measurable criteria or mark for clarification.
Modes
This skill has two modes — pick by user intent or orchestrator parameter:
- specify (default) — write a new spec or major rewrite
- clarify — resolve
[NEEDS CLARIFICATION]markers in an existing spec
Workflow — specify mode
Step 1 — Read existing context
In priority order:
docs/constitution.md— required. If missing, offerproject-constitutionfirst.docs/product-soul.md— strategic grounding (optional).docs/prd/<latest>.md— if a PRD exists, import problem framing and user context.docs/specs/<latest>-design.md— ifbrainstormingproduced a design doc, import the approach (but discard architecture sections).
Step 2 — Discovery (max 3 questions, one at a time)
Ask only what cannot be inferred:
- "What is the user-visible outcome when this works?"
- "What are the 2–3 most important things this MUST NOT do (out of scope)?"
- "Are there constitutional rules this feature has to specifically address?"
If the request is too vague to draft FRs, mark them
[NEEDS CLARIFICATION]and continue — don't loop in interview.
Step 2b — Reframe vague requirements
Adjectives ("fast", "intuitive") → measurable criteria (latency, error rate, completion %) — confirm targets with user before drafting FRs.
Step 3 — Write the spec
Before drafting, if any requirement is inferred (stack, auth model, deployment target), list up to 5 bullets under ## Assumptions I'm Making and ask the user to confirm or correct — do not silently fill gaps.
Read references/feature-spec-schema.md for the full template. Required sections:
- Frontmatter (artifact, status, constitution version, sources, slug)
- Summary (1–2 sentences)
- Problem
- User Scenarios (US-1, US-2, …)
- Functional Requirements (FR-1, FR-2, …)
- Non-Functional Requirements (NFR-1, NFR-2, …)
- Acceptance Criteria (AC-FR-1.1 in Given/When/Then form — written so each AC converts mechanically to a failing-test skeleton: Given→arrange, When→act, Then→assert; see
references/feature-spec-schema.md→ Test Skeletons) - Edge Cases (minimum 3)
- Out of Scope
- Constitution Waivers (only if any rule is intentionally not satisfied)
- Needs Clarification (CL-1, CL-2, …)
- Review Checklist
Set status: Draft if any clarifications remain, status: Clarifying while user is resolving, status: Approved only when CL list is empty AND user explicitly approves.
Step 4 — Self-review
- No HOW (no libraries, file paths, architecture, code patterns)
- Every FR/NFR has at least one AC in Given/When/Then form
- Out of Scope is non-empty and specific
- No vague adjectives ("fast", "easy", "intuitive")
- Every unresolved question is in
Needs Clarificationwith a CL ID - Constitution version referenced; any waivers are explicit
Step 5 — Save, log, notify
Save to: docs/specs/YYYY-MM-DD-<slug>-feature-spec.md
Append to docs/skill-outputs/SKILL-OUTPUTS.md:
| YYYY-MM-DD HH:MM | feature-spec | docs/specs/YYYY-MM-DD-<slug>-feature-spec.md | Spec: <title> (status) |Tell the user:
"Feature spec saved (status:
<status>).<N>clarifications remain — run me in clarify mode to resolve them, or invokespec-driven-development /clarify. Once Approved, ask me to emit failing-test skeletons from the ACs —/implementstarts red from them."
Step 6 — Memory Checkpoint (Mandatory)
Per memory/SKILL.md → Mandatory Auto-Trigger Checkpoints (event: feature-spec written), invoke memory-capture with spec slug, status, and key requirements/constraints for next-agent continuity.
Workflow — clarify mode
Step C1 — Load the spec
Read the named (or latest) docs/specs/*-feature-spec.md.
Step C2 — Walk clarifications one at a time
For each CL-N:
- Show the question with surrounding context.
- Wait for user answer.
- Update the relevant FR/NFR/AC. Replace the
[NEEDS CLARIFICATION]marker with the answer. - Remove
CL-Nfrom Needs Clarification list.
Step C3 — Update status (HYPOTHESIS + CONFIDENCE %)
After every answered CL, record internally:
- HYPOTHESIS: one sentence — what the spec now says about this CL.
- CONFIDENCE: integer % (0–100) the resolved FR/NFR/AC is unambiguous enough for
spec-crosscheckPASS and an implementing agent to plan from with zero follow-up.
Promote to Clarifying-Complete only when every resolved CL has CONFIDENCE ≥ 70%. Any CL <70% gets a one-line REASON and is re-opened as CL-N (revisit) rather than silently closed.
When CL list is empty AND all resolutions ≥70%, ask:
"All clarifications resolved (avg confidence: N%). Approve as final? (yes → status: Approved)"
Only set Approved after explicit user confirmation.
Step C4 — Save, log, notify
Re-save to the same path. Append:
| YYYY-MM-DD HH:MM | feature-spec | <path> | Spec clarified: <title> → <status> |Gotchas
- "WHAT not HOW" is the bright line. If you catch yourself writing "use Postgres" or "in
services/auth.ts" — stop. That belongs inimplementation-plan. - Acceptance criteria must be testable as written. "Login works" fails. "Given valid credentials, When user submits, Then JWT is returned within 500ms" passes. Litmus: if an AC cannot become a failing-test skeleton as written, it is not done.
- Edge cases are not nice-to-have — they're how
spec-crosscheckdetects missing tasks. Brainstorm at least 3. Out of Scopemust be specific — strongest anti-scope-creep tool.- Constitution waivers need explicit
## Constitution Waiverswith rule ID + rationale.
Example
<examples>
<example>
<input>/specify a feature for password-less email magic-link login.</input>
<output>
Reading `docs/constitution.md@2` (auth rule C-2.4: tokens MUST expire ≤ 15 min).Spec drafted with:
- 2 user scenarios
- 5 FRs (request link, validate token, single-use enforcement, expiry, rate limit)
- 3 NFRs (latency budget per AC, GDPR consent, audit log)
- 7 ACs in Given/When/Then form
- 3 edge cases (expired token, replay, multiple devices)
- Out of scope: SSO, OAuth, social login
- 2 [NEEDS CLARIFICATION]:
- CL-1: Token TTL within constitution limit — 5, 10, or 15 minutes?
- CL-2: Should a second click on a used link return generic 404 or "already used"?
Saved as docs/specs/2026-05-02-magic-link-feature-spec.md (status: Draft).
Run /clarify next.
</output>
</example>
</examples>
Common Rationalizations
| Excuse | Reality |
| Spec can include HOW | WHAT/WHY only — HOW belongs in implementation-plan. |
| Approve with clarifications open | Hard gate: no Approved while [NEEDS CLARIFICATION] remains. |
| Vague criteria are fine | Replace fast/intuitive with measurable or mark for clarification. |
Verification
- Constitution version referenced or gap explicit
- FR/NFR/AC complete; no open clarification markers at Approved
- No implementation details in spec body
- spec-crosscheck can trace every requirement
Red Flags
- Spec drifts into HOW — stack or file paths in requirements
- Acceptance criteria not testable as written
- Edge cases omitted that block spec-crosscheck coverage
- Needs Clarification list left non-empty at approval
Prune Log
Last pruned: 2026-07-09
- Added AC→failing-test-skeleton contract (Given→arrange, When→act, Then→assert) + schema Test Skeletons section (agent-loom Phase 5, SDD×TDD)
Impact Report
Feature spec: <title> Status: Draft | Clarifying | Approved Constitution: docs/constitution.md@<N> Counts: US=<N> FR=<N> NFR=<N> AC=<N> Edge=<N> CL=<N> Saved: docs/specs/YYYY-MM-DD
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.

