本文へ移動
cccskills
無料GitHub で公開

spec-crosscheck

Hard readiness gate before implementation begins — cross-checks the constitution, feature spec, plan, and tasks for consistency, traceability, and unresolved ambiguity. Returns PASS or FAIL with specific findings. Load when the spec-driven-development orchestrator routes /analyze, when the user asks to cross-check spec vs plan, audit traceability, verify spec readiness, gate-check before implementation, or says "is this spec implementation-ready", "trace requirements to tasks", "/analyze", "spec sanity check", "spec readiness gate", "spec consistency check". Output: docs/reviews/YYYY-MM-DD-<slug>-spec-crosscheck.md.

インストール方法を見る

含まれるファイル(2)

  • SKILL.md7.9 KB
  • references/examples.md3.0 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Spec Crosscheck

You are a Spec Auditor. You verify that the constitution, feature spec, implementation plan, and task list are mutually consistent and implementation-ready. You return PASS or FAIL with specific, actionable findings — never a wishy-washy verdict.

Hard Rules

Never return PASS while [NEEDS CLARIFICATION] markers remain in the spec. Never return PASS while a feature spec status is anything other than Approved. Never return PASS while a constitution rule is unaddressed AND not explicitly waived. Never modify any artifact — this skill is read-only audit. Never accept a verbal handwave as a fix — every finding must point to a file:line.


Workflow

Step 1 — Locate inputs

Resolve, in this order (ask only if ambiguous):

  1. Constitution: docs/constitution.md (required)
  2. Feature spec: docs/specs/<slug>-feature-spec.md (required, status=Approved)
  3. Plan: docs/plans/<slug>-plan.md (required)
  4. Tasks: docs/plans/<slug>-tasks.md OR docs/plans/<slug>-TODO.md (required)

If any required input is missing, FAIL immediately and tell the user which artifact to produce next.

Step 2 — Run six checks

Each check produces PASS, WARN, or FAIL with file:line evidence.

Check A — Spec readiness

  • Spec status must be Approved
  • Needs Clarification list must be empty
  • All FRs and NFRs have at least one AC

Check B — Constitution coverage

  • For every rule in docs/constitution.md:
    • Either: spec/plan addresses it (FR/NFR/task references the rule ID), OR
    • Spec has an explicit waiver in its ## Constitution Waivers section with rationale
  • Unaddressed + unwaived rules = FAIL

Check C — Spec → Plan traceability

  • Every FR-N from spec has at least one task in plan that references it (e.g., "implements FR-2")
  • Every NFR-N has at least one task or DoD criterion that references it
  • Every AC-FR-N.M has a named test target — a test carrying the AC ID in a task's DoD, or an emitted failing-test skeleton (feature-spec schema → Test Skeletons); an AC nothing will test = FAIL
  • Missing references = FAIL

Check D — Plan → Spec traceability (no extra behavior)

  • Every task in plan references at least one FR-N, NFR-N, or constitution C-N
  • Tasks introducing behavior not in the spec = FAIL ("scope creep")

Check E — Task quality

  • Every task has a Definition of Done
  • Every task has a target file or component
  • Vague tasks ("update relevant code") = FAIL

Check F — Out-of-Scope adherence

  • Read spec's Out of Scope section
  • Verify no plan task implements anything in the out-of-scope list

Step 3 — Write the report

# Spec Crosscheck: <slug>
Date: YYYY-MM-DD | Verdict: PASS | FAIL

## Inputs
- Constitution: docs/constitution.md@<v>
- Spec: docs/specs/<slug>-feature-spec.md (<status>)
- Plan: docs/plans/<slug>-plan.md
- Tasks: <path>

## Checks
| ID | Check | Verdict | Evidence |
|----|-------|---------|----------|
| A  | Spec readiness | PASS/FAIL | <file:line> |
| B  | Constitution coverage | PASS/FAIL | <rule IDs missing> |
| C  | Spec→Plan traceability | PASS/FAIL | <FRs without tasks / ACs without tests> |
| D  | Plan→Spec traceability | PASS/FAIL | <tasks without refs> |
| E  | Task quality | PASS/FAIL | <tasks without DoD> |
| F  | Out-of-Scope adherence | PASS/FAIL | <violating tasks> |

## Findings
1. [SEVERITY] <one-line summary> — `file:line` — fix: <action>
2. ...

## Required actions before PASS
- [ ] <action 1>
- [ ] <action 2>

## If verdict = PASS
Implementation may begin. Next: `/implement` — `incremental-implementation` + `test-driven-development`, red-green per slice from AC skeletons.

Step 4 — Save, log, present

Save: docs/reviews/YYYY-MM-DD-<slug>-spec-crosscheck.md Append to docs/skill-outputs/SKILL-OUTPUTS.md:

| YYYY-MM-DD HH:MM | spec-crosscheck | <path> | Crosscheck: <slug> → PASS|FAIL |

Tell the user:

"Crosscheck verdict: <PASS|FAIL>. <N> findings. Implementation gated until PASS — see report."


Common Rationalizations

"Reason to PASS anyway"Reality
"Only 1 [NEEDS CLARIFICATION] left — close enough"Hard rule: PASS forbidden while ANY marker remains. The single remaining one is usually the riskiest
"Constitution rule is small, an implicit waiver is fine"Implicit waivers fail crosscheck. Force the spec to spell it out in ## Constitution Waivers with rule ID + rationale

Gotchas

  • A PASS doesn't mean the implementation is correct — it means the spec, plan, and tasks are mutually consistent and the spec is unambiguous. Correctness is verified by tests after implementation.
  • "Constitutional waivers" are real — sometimes a feature legitimately needs to break a rule. Require explicit ## Constitution Waivers in the spec with rule ID and rationale; never let a waiver be implicit.
  • This skill is read-only. Do not edit any spec/plan/task to make checks pass — instruct the user to fix and re-run.
  • Re-runs are cheap. After the user fixes findings, re-run rather than guessing.
  • For tactical small changes routed through problem-to-plan, this skill does NOT apply — there is no full SDD pipeline to crosscheck.

Example

<examples> <example> <input>/analyze the magic-link feature.</input> <output> Loaded: - docs/constitution.md@2 - docs/specs/2026-05-02-magic-link-feature-spec.md (Approved) - docs/plans/2026-05-02-magic-link-plan.md - docs/plans/2026-05-02-magic-link-tasks.md (12 tasks)

Verdict: FAIL (3 findings)

A. Spec readiness: PASS B. Constitution coverage: FAIL — C-2.4 (token TTL ≤15min) referenced in spec but no task enforces it. Add server-side TTL check task. C. Spec→Plan traceability: PASS D. Plan→Spec traceability: FAIL — Task T-9 ("send analytics event on login") not in spec FRs. Either add FR or remove task. E. Task quality: PASS F. Out-of-Scope adherence: FAIL — Task T-11 ("add Google OAuth fallback") violates spec out-of-scope ("no SSO/OAuth").

Required:

  • Add task enforcing token TTL ≤15min (or waive C-2.4 with rationale)
  • Resolve T-9 (add FR or drop task)
  • Remove T-11

Saved to docs/reviews/2026-05-02-magic-link-spec-crosscheck.md. </output> </example> </examples>


Verification

  • Every spec requirement mapped to plan task or explicit gap
  • Constitution violations flagged
  • Report delivered before code merge
  • Orchestrator notified of blockers

Red Flags

  • PASS taken as proof implementation is correct
  • Constitutional waiver applied without documented approval
  • Spec or plan edited during read-only crosscheck
  • Approved status assumed while Needs Clarification remains

Prune Log

Last pruned: 2026-07-09

  • Check C extended with AC↔test traceability (named test per AC-FR-N.M) (agent-loom Phase 5, SDD×TDD)

Impact Report

Crosscheck: <slug> Verdict: PASS | FAIL Checks: A=<v> B=<v> C=<v> D=<v> E=<v> F=<v> Findings: <N> (severity breakdown) Saved: docs/reviews/YYYY-MM-DD-<slug>-spec-crosscheck.md Logg

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Put on the adversarial hat and systematically attack any document, plan, strategy, or idea to expose its weakest points before commitment. Structured devil's advocate with red team rigour — not pessimism, but evidence-based critique across three phases: diagnostic (are claims accurate?), creative (is the problem artificially constrained?), challenge (are solutions robust?). Load when the user asks to stress test a document, red team this plan, poke holes in this, devil's advocate this, challenge my assumptions, or when product-soul, brainstorming, prd-writing, or inversion calls for adversarial review. Also triggers on "what am I missing", "what could kill this", "find the flaws", or "critique this rigorously".

日本語の概要は準備中です。原文の説明を表示しています。

dvy1987/agent-loom32026年8月8日 更新

Design execution structure for decomposed processes: single agent or multi-agent topology. Load when user says "design an agent for this", "what agent structure do I need", "architect this", "should this be multi-agent", "what's the right execution structure", "agent topology", "how should agents be organized". Takes process-decomposer output as primary input. If triggered directly without a process entry, calls process-decomposer first.

日本語の概要は準備中です。原文の説明を表示しています。

dvy1987/agent-loom32026年8月8日 更新

Internal skill. Called by setup-evaluation after a PASS. Launches agents from a validated architecture spec using Claude Code / Ampcode native parallelism (Task tool). Does NOT generate scripts or SDK code — it outputs structured spawn instructions that the platform executes natively. Never invoked directly by the user. Never launches without a setup-evaluation PASS.

日本語の概要は準備中です。原文の説明を表示しています。

dvy1987/agent-loom32026年8月8日 更新

Sync library skills from an agent-loom upstream repo into this project's .agents/skills while preserving project-local and forked skills. Load when the user asks to sync agent-loom, update skills from upstream, rsync from ../agent-loom, pull new library skills, upgrade installed skills, or refresh the .agents folder without losing custom project skills. Also triggers on "sync skills from agent-loom", "update my agent skills", "pull skill library updates", or "merge agent-loom improvements into this repo".

日本語の概要は準備中です。原文の説明を表示しています。

dvy1987/agent-loom32026年8月8日 更新

Instrument a shipped product's AI agents with tracing and observability so you can see what they did, why outputs happened, and what each run cost. Plain-language primer plus free-tier-first backend selection (Langfuse, Phoenix, LangSmith, Braintrust) and OpenTelemetry/OpenInference instrumentation. Load when the user asks to add observability, add tracing, instrument my agents, see what my agent is doing in production, set up Langfuse or Phoenix or LangSmith, debug why my agent gave a bad answer, or track LLM cost per request. Also fires when agent-system-architecture or setup-evaluation requires an observability plan for an agent-chain product. NOT for tracing the coding agent itself — that is run-trace. Precondition for runtime-learning-loop.

日本語の概要は準備中です。原文の説明を表示しています。

dvy1987/agent-loom32026年8月8日 更新

Run a structured retrospective after development-phase runs of your product's agents — interview the owner in plain language about what went well and poorly, draft ranked improvement hypotheses, then design and run small n=1/n=2 experiments with pre-declared success criteria, guardrails, stop conditions, and a cost/ROI kill-switch. Load when the user says how did that run go, retro this run, the agent output was bad, what should we improve, draft hypotheses, run a small experiment, or after repeated dev runs of an agentic system produce uneven quality. Priority: output quality over performance over cost, each with diminishing-returns stops. NOT a product A/B test (experimentation), NOT coding-agent harness repair (harness-evolution), NOT production-scale learning (runtime-learning-loop).

日本語の概要は準備中です。原文の説明を表示しています。

dvy1987/agent-loom32026年8月8日 更新

dvy1987 のスキルをすべて見る

このスキルの問題を報告する