AgentSkillsCN

validation-report

验证报告的格式与测试规范,用于依据规格验证实施效果。当您规划、执行或汇报运行时对规格的验证工作时,可使用此技能。

SKILL.md
--- frontmatter
name: validation-report
description: Validation report format and testing conventions for verifying implementation against spec. Use when planning, executing, or reporting validation of the runtime against its specification.

Validation Report — Skill

Purpose

Define the format and approach for independently validating that the implementation satisfies all spec requirements and scenarios.

When to Use

  • Planning a validation phase
  • Executing validation tests
  • Writing the final validation report

Output Template

markdown
# VALIDATION.md — [Project Name]

## 1. Validation Summary
- **Date**: [date]
- **Spec version**: [commit hash of SPEC.md]
- **Implementation version**: [commit hash]
- **Verdict**: PASS | FAIL | PARTIAL

## 2. Environment
- **Node.js version**: [version]
- **OS**: [os]
- **Setup steps**: [how environment was prepared]

## 3. Automated Test Results
- **Command**: `npm test`
- **Exit code**: [0 or non-zero]
- **Output**: [full test output]

## 4. Scenario Coverage Matrix

| Scenario | Spec Section | Requirements | Result | Evidence |
|----------|-------------|-------------|--------|----------|
| [name]   | 6.N         | REQ-XX-NNN  | PASS/FAIL | [output excerpt or test name] |

## 5. Requirement Coverage Matrix

| Requirement | Scenarios Covering It | All Pass? |
|-------------|----------------------|-----------|
| REQ-XX-NNN  | 6.N, 6.M            | Yes/No    |

## 6. Issues Found
[List any failures, with: what failed, expected vs actual, severity, affected requirements.]

## 7. Verdict Rationale
[Evidence-based summary explaining the verdict. Cite specific scenario results.]

Conventions

  1. Clean environment. Validation starts from a fresh state — no inherited runtime process, no cached modules.
  2. Automated first. Run npm test before any manual checks. The automated suite is the primary evidence.
  3. Every spec scenario (Section 6) must be covered. If a scenario is not tested by the automated suite, test it manually and document the result.
  4. Every requirement must appear in the coverage matrix. Requirements not covered by any scenario are flagged.
  5. Evidence is specific. "It works" is not evidence. Cite test names, HTTP status codes, response bodies, and exact output.
  6. Verdict is conservative. Any failing scenario means FAIL unless the failure is cosmetic and all requirements are still met.
  7. Use different test data than implementation demos where possible. Validation should not just replay the happy path the developer tested.

Quality Checklist

  • All 12 spec scenarios (6.1–6.12) have a row in the coverage matrix
  • All 18 requirements have a row in the requirement coverage matrix
  • npm test output is included verbatim
  • Every FAIL result has a specific issue description
  • Verdict matches the evidence (no optimistic overrides)
  • Environment section is complete and reproducible