スキル一覧に戻る
doodledood

verify

by doodledood

A curated marketplace of Claude Code plugins for agentic development workflows, featuring tools for architecture, knowledge management, and development automation

8🍴 0📅 2026年1月24日
GitHubで見るManusで実行

SKILL.md


name: verify description: 'Manifest verification runner. Spawns parallel verifiers for Global Invariants and Acceptance Criteria. Called by /do, not directly by users.' user-invocable: false

/verify - Manifest Verification Runner

Goal

Run all verification methods from a Manifest. Spawn one verifier agent per criterion in parallel. Report results grouped by type.

Input

$ARGUMENTS = " [--scope=files]"

Principles

  1. Don't run checks yourself - Spawn agents to verify. You orchestrate, they verify.

  2. Single parallel launch - All criteria in one call, slow ones first (tests, builds, reviewers before lint/typecheck).

  3. Global failures are critical - Highlight prominently. Task can't succeed while these fail.

  4. Actionable feedback - Pass through file:line, expected vs actual, fix hints.

What to Do

Parse inputs - Extract all criteria with verification methods from manifest. Read execution log for context.

Categorize by method:

  • bash: Shell commands (tests, lint, typecheck)
  • subagent: Reviewer agents
  • codebase: Code pattern checks
  • manual: Set aside for human verification

Launch verifiers - One Task per criterion, all in parallel. Use the agent type specified in the manifest's verification block. Pass criterion ID, description, verification method, and relevant context.

Collect and report results - Group by Global Invariants first, then by Deliverable.

Decision Logic

if any Global Invariant failed:
    → Return ALL failures, globals highlighted prominently

elif any AC failed:
    → Return failures grouped by deliverable

elif all automated pass AND manual exists:
    → Return manual criteria, hint to call /escalate

elif all pass:
    → Call /done

Output Format

On failure:

## Verification Results

### Global Invariants

#### Failed (N)
- **INV-G1**: [description]
  Method: [method]
  [failure details with location, expected/actual, fix hint]

#### Passed (M)
- INV-G2, INV-G3

---

### Deliverable 1: [Name]

#### Failed
- **AC-1.2**: [description]
  [failure details]

#### Passed
- AC-1.1

---

**Summary:**
- Global Invariants: X/Y failed (fix first)
- Deliverable 1: A/B ACs failed

On success with manual:

## Verification Results

All automated criteria pass.

### Manual Verification Required
- **AC-1.3**: [description] - How to verify: [from manifest]

Call /escalate to surface for human review.

On full success: Call /done.

Criterion Types

TypePatternFailure Impact
Global InvariantINV-G{N}Task fails
Acceptance CriteriaAC-{D}.{N}Deliverable incomplete
Process GuidancePG-{N}Not verified (guidance only)

Note: PG-* items are non-verifiable guidance on HOW to work. They're followed during /do but not checked by /verify.

スコア

総合スコア

60/100

リポジトリの品質指標に基づく評価

SKILL.md

SKILL.mdファイルが含まれている

+20
LICENSE

ライセンスが設定されている

0/10
説明文

100文字以上の説明がある

+10
人気

GitHub Stars 100以上

0/15
最近の活動

3ヶ月以内に更新がある

0/10
フォーク

10回以上フォークされている

0/5
Issue管理

オープンIssueが50未満

+5
言語

プログラミング言語が設定されている

+5
タグ

1つ以上のタグが設定されている

0/5

レビュー

💬

レビュー機能は近日公開予定です