test-case-reviewer-plus

test-case-reviewer-plus (EN)

Chinese version： See the corresponding Chinese skill.

When to Use

• Need a stricter review of existing test cases before test execution or release.
• Need issue severity, business impact, and retest order to be explicit.

Script capability boundary

Optional scripts/run_*.py prepare parsed evidence or source candidates with NOT_ASSESSED. They do not call a model, execute the main prompt, or produce semantic conclusions, risk priorities, complete cases, or approval. An agent performs the main prompt. Read references/script-contract.md before running helpers.

Workflow

1. Read and follow the main prompt listed under Progressive disclosure (coverage, structure, quality bar).
2. Accept direct materials or user-supplied multi-role review reports; when role reports are used, retain source_role and source identifiers.
3. Add only project context that changes the result: scope, environment, constraints, risks, dependencies, expected deliverable.
4. If input is incomplete, return a usable first draft and explicitly mark assumptions and gaps.
5. Default to Markdown; switch formats only when the user asks.

Core Constraints

• Prioritize by risk / business impact — do not treat everything equally.
• Separate confirmed facts from current assumptions.
• Do not invent endpoints, fields, environments, or root causes the user did not provide.
• Keep output executable: concrete scenarios, clear priority, clear next steps.
• Keep blockers, high-risk coverage gaps, maintainability findings, and low-value or duplicate cases in distinct sections rather than one generic issue list.
• Produce only AI findings and a recommendation; always output human_final_decision: pending, with the final pass, conditional-pass, or reject decision owned by a Human.

Progressive Disclosure

• Before producing output, read and follow prompts/test-case-reviewer-plus.md (minimum coverage, output structure, quality bar).
• When a ready-made template fits: use matching files under output-templates/.
• When the user wants examples or alignment with existing assets: read relevant examples/.
• For format conversion or helper checks: prefer existing scripts/ over reinventing.
• For evaluating/regressing this skill: use evals/ with skill-up.

Pre-delivery Checklist

• Followed the main prompt's output structure
• Minimum coverage focus: high-severity findings, coverage gaps, missing positive scenarios, missing negative scenarios, missing boundary scenarios, traceability, step and expectation quality, business impact, ... (details in main prompt)
• Covered the minimum checklist, or explained omissions
• High-risk items have explicit priority
• Did not invent details the user did not provide
• Assumptions and gaps are marked
• Multi-role inputs retain role sources and the four key finding categories remain distinct
• The AI recommendation does not impersonate a final Human approval or rejection record

Common Pitfalls

• Do not pretend completeness when scope/context is missing.
• Do not treat every item as equally important.
• Do not skip assumptions and information gaps.
• Do not turn an AI pass/reject recommendation into a final Human decision.
• Do not dump generic theory unrelated to the current toolchain.