본문으로 건너뛰기
cccskills
무료GitHub에 공개

adherence-check

Check any artifact against an explicit specification and report adherence using the add-comments workspace format. Use when validating plans, scripts, docs, or code changes against a source-of-truth spec.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

설치 방법 보기

포함된 파일(1)

  • SKILL.md5.5 KB

SKILL.md(원문)

설치하기 전에 에이전트에게 전달되는 지시 내용을 확인할 수 있습니다.

Adherence Check

Purpose

Verify that a specific artifact adheres to a given specification, then report any violations using the add-comments workspace format.

This skill is repo-agnostic. It works in any repository as long as you supply the three required inputs.

Inputs (All Required)

InputDescription
artifact_pathPath to the file or artifact to check
spec_pathPath to the source-of-truth specification to check against
report_pathPath to the file where the adherence report will be appended

If any input is missing, STOP and ask the caller to provide it before proceeding.

Prerequisite Gate

Before starting evaluation, verify that both artifact_path and spec_path resolve to existing, readable files.

  • If spec_path is missing or the file does not exist: STOP immediately. Report the missing prerequisite to the caller. Do not improvise a specification. Do not guess requirements.
  • If artifact_path is missing or the file does not exist: STOP immediately. Report the missing artifact to the caller.

Procedure

Step 1: Read the Specification

  1. Read the file at spec_path.
  2. Identify the set of requirements, rules, or constraints it defines.
  3. If the spec contains a quick-start matrix, change-type index, or table of contents, use it to scope which sections apply to the artifact. Record the list of applicable sections (numbers and titles, if available).
  4. If the spec has no such index, treat every requirement as potentially applicable.

Step 2: Read the Artifact

Read the artifact at artifact_path. Identify the specific behaviors, statements, or instructions that must align with the spec requirements.

For large artifacts:

  • Scan headings or file structure first.
  • Deep-read only the sections relevant to the applicable requirements.
  • For code artifacts, focus on behavior and implications; cite file paths and line numbers where possible.

Step 3: Evaluate Each Requirement

For each applicable requirement in the spec:

  • PASS -- the artifact clearly satisfies the requirement.
  • FAIL -- the artifact violates or omits the requirement.
  • WARN -- the requirement may be violated but evidence is ambiguous; explain the ambiguity. When unsure whether something is a true violation, use WARN rather than FAIL.
  • N/A -- the requirement does not apply to this artifact.

Record FAIL and WARN items in the report details. Include all counts in the summary.

Tip: Keep a simple tally table as you go (PASS / FAIL / WARN / N/A) to avoid count errors.

Step 4: Write the Report (Exact Structure)

Draft the report in a scratch buffer or temporary file first, then append via add-comments.

Use this structure exactly:

# Adherence Check Report

**Artifact checked**: [artifact_path]
**Spec checked against**: [spec_path]
**Checked by**: [Agent name]
**Timestamp**: [YYYY-MM-DD HH:MM:SS]
**Artifact type**: [e.g., script, plan, doc, code module]
**Sections checked**: [List section numbers + titles, or "all" if spec has no sections]

## Summary
- Requirements checked: [N]
- PASS: [X]
- FAIL: [Y]
- WARN: [W]
- NOT APPLICABLE: [Z]

## Violations

### FAIL: [Requirement identifier] - [Requirement title or short description]

**Requirement**: [Quote or paraphrase the spec requirement]

**Artifact evidence**: [Quote exact text or cite line numbers / snippet]

**Gap**: [Specific mismatch]

**Suggested fix**: [Concrete correction]

---

### WARN: [Requirement identifier] - [Requirement title or short description]

**Requirement**: [Quote or paraphrase the spec requirement]

**Artifact evidence**: [Quote exact text or cite line numbers / snippet]

**Gap**: [Specific mismatch + why it is ambiguous]

**Suggested fix**: [Concrete correction or clarification needed]

---

[Repeat for each failure or warning]

## Notes (Optional)
- [Any brief caveats or follow-ups]

If there are no violations:

## Violations

None found.

Step 5: Deliver via add-comments

Use the add-comments skill to append the report to report_path:

  1. Add or locate the main header: # AI AGENT COMMENTS AND WORK FOLLOW
  2. Add or locate your workspace header: ## Codex Workspace
  3. Append a timestamped entry and include the report.
  4. End with ---.

Do not edit any other agent workspaces. Never overwrite or replace the file; append only via add-comments. If the add-comments skill is missing or broken, STOP and report the dependency failure.

Completion Checklist

  • Both spec_path and artifact_path verified to exist
  • Spec read and requirements cited by section/identifier
  • Artifact read and referenced with evidence
  • Report matches the exact structure above
  • Each FAIL/WARN includes Requirement, Evidence, Gap, Suggested fix
  • Report delivered using add-comments workspace rules
  • Summary math is correct (PASS + FAIL + WARN + N/A = Requirements checked)

Example Invocation

"Run adherence-check.
  artifact_path: src/deploy.sh
  spec_path: docs/deploy-spec.md
  report_path: reports/adherence-results.md"

리뷰

아직 리뷰가 없습니다. 사용해 본 소감을 남겨 주세요.

댓글

로그인 없이 쓸 수 있습니다. 사용 후기, 질문, 작동하지 않은 환경 등을 편하게 남겨 주세요.

아직 댓글이 없습니다. 첫 댓글을 남겨 보세요.

보내면 바로 공개됩니다. 이메일 주소나 전화번호 같은 개인 정보는 쓰지 마세요. URL은 2개까지입니다.

0 / 1,000자

이 양식은 reCAPTCHA로 보호되며 Google 개인정보처리방침 및 서비스 약관이 적용됩니다.

같은 저장소의 스킬

Method for adding agent comments and work to a shared file without overlapping other agents. Creates isolated workspaces per agent with UTC timestamps. Use when told to add comments, reviews, or work to any file.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

dazzaji/interlateral_agents112026년 9월 28일 업데이트

Join the Antigravity CLI (agy, Gemini 3.5 Flash) to the Interlateral mesh as a native CLI peer with its own tmux session, identity stamping, and the agy.js send helper. Use this instead of the Antigravity desktop-app CDP path.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

dazzaji/interlateral_agents112026년 9월 28일 업데이트

Agents work in parallel on the same task; judges evaluate and select the winner.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

dazzaji/interlateral_agents112026년 9월 28일 업데이트

concierge

무료

Boot the explicitly selected CLI(s) as the human's concierge: a project-aware side channel that sits ABOVE the running agent system in a shared writable terminal. It answers status questions and handles the human's ad hoc tasking while the working agents collaborate on the mesh — it is not a worker, manager, or gatekeeper, and team progress never waits for it.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

dazzaji/interlateral_agents112026년 9월 28일 업데이트

Multiple agents draft a structured document through federated co-authorship and formal voting.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

dazzaji/interlateral_agents112026년 9월 28일 업데이트

All agents have equal voice; decisions made by majority vote.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

dazzaji/interlateral_agents112026년 9월 28일 업데이트

dazzaji의 스킬 모두 보기

이 스킬의 문제 신고