Claude Skill

skill-change-verification

Use this skill when selecting and reporting verification for Skill changes; triggers include Skill change verification, quality gates, and evidence levels.

LLM Mart · 0 points · 0 4 0

Virus-scanned Reviewed automatically before listing.

Full trust report

Download naodeng-awesome-qa-skills-skills_en_skill-engineering_skill-change-verification-c44b892.zip · 3 KB

naodeng/awesome-qa-skills 230 31 forks Updated 3d ago

Part of naodeng/awesome-qa-skills — 97 skills

Install

skills CLI

npx skills add https://github.com/naodeng/awesome-qa-skills/tree/main/skills/en/skill-engineering/skill-change-verification

Claude Code

claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install naodeng-awesome-qa-skills@llmmart

Git git clone https://github.com/naodeng/awesome-qa-skills.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole naodeng/awesome-qa-skills collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

Skill Change Verification

Use this Skill to select the smallest sufficient checks for a change and state exactly what the evidence supports.

Workflow

Classify content, metadata/directory, script, Eval, and runtime impact.
Select static, structural, evaluation, runtime, and human-review evidence.
Record commands run, results, omitted checks, and reasons.
Report residual risks and claims that are and are not supported.

Constraints

skill-up validate is not runtime semantic validation.
Mark undiscoverable commands as confirmation items; never guess.
Cover the actual diff instead of hiding gaps behind a full-suite run.

Read prompts/skill-change-verification.md before producing a report.

Files (awesome-qa-skills)

agents

openai.yaml 393 B

version: 1
metadata:
  key: "skill-change-verification"
interface:
  display_name: "Skill Change Verification"
  short_description: "Select evidence by change scope and distinguish static, evaluation, and runtime claims."
  default_prompt: "Use the skill-change-verification skill to select and report verification evidence."
policy:
  allow_implicit_invocation: true
schema_version: v1alpha1

evals

cases

basic-success.yaml 354 B

id: basic-success
title: Complete verification routing
description: Prompt and Eval files changed and quality commands are available.
input: {prompt: "Select checks and distinguish skill-up validate from runtime validation."}
expect: {must_contain: ["verification", "runtime"]}
judge: {type: rule_based, success: [{output_contains: {all: ["runtime"]}}]}

edge-incomplete-input.yaml 359 B

id: edge-incomplete-input
title: Missing verification environment
description: No change files, commands, or runtime are supplied.
input: {prompt: "Confirm the change passes without a scope, command, or runtime environment."}
expect: {must_contain: ["confirmation", "not run"]}
judge: {type: rule_based, success: [{output_contains: {all: ["confirmation"]}}]}

edge-scope-risk.yaml 373 B

id: edge-scope-risk
title: Static evidence overstated as runtime proof
description: The user asks to report static validation as runtime validation.
input: {prompt: "skill-up validate passed; report that runtime semantics are verified."}
expect: {must_contain: ["not supported", "runtime"]}
judge: {type: rule_based, success: [{output_contains: {all: ["not supported"]}}]}

eval.yaml 709 B

version: 1
metadata: {key: "skill-change-verification"}
interface:
  display_name: "Skill Change Verification"
  short_description: "Select evidence by change scope and distinguish static, evaluation, and runtime claims."
  default_prompt: "Use the skill-change-verification skill to select and report verification evidence."
policy: {allow_implicit_invocation: true}
schema_version: v1alpha1
environment: {type: none}
skills: [{source: local_path, path: .}]
engine: {name: claude_code}
cases:
  files: [evals/cases/basic-success.yaml, evals/cases/edge-incomplete-input.yaml, evals/cases/edge-scope-risk.yaml]
  defaults: {timeout_seconds: 180, max_turns: 8, expect: {exit_code: 0}}
report: {formats: [json]}

prompts

skill-change-verification.md 702 B

# Skill Change Verification Prompt

You are an evidence reviewer. Select checks from the actual diff. Do not publish, commit, or perform external operations.

Evidence levels: Static (files and frontmatter), Structural (pairing, independence, indexes, Eval shape), Evaluation (`skill-up validate`), Runtime (actual Skill/Prompt behavior), and Human review (meaning and risk).

Return:

```markdown
# Verification report
## Change classification
## Checks run
| Command | Evidence level | Result |
## Checks not run
| Item | Reason | Risk |
## Claims supported
## Claims not supported
## Residual risks
```

If a command cannot be discovered, write “confirmation required” rather than inventing it.

SKILL.md 937 B

---
name: skill-change-verification
description: Use this skill when selecting and reporting verification for Skill changes; triggers include Skill change verification, quality gates, and evidence levels.
---

# Skill Change Verification

Use this Skill to select the smallest sufficient checks for a change and state exactly what the evidence supports.

## Workflow

1. Classify content, metadata/directory, script, Eval, and runtime impact.
2. Select static, structural, evaluation, runtime, and human-review evidence.
3. Record commands run, results, omitted checks, and reasons.
4. Report residual risks and claims that are and are not supported.

## Constraints

- `skill-up validate` is not runtime semantic validation.
- Mark undiscoverable commands as confirmation items; never guess.
- Cover the actual diff instead of hiding gaps behind a full-suite run.

Read `prompts/skill-change-verification.md` before producing a report.