When to Use
Use when you need to select the minimal but sufficient checks based on the scope of change, and accurately state what conclusions the evidence can support.
Execution Flow
- Classify the change into content, metadata/directory, scripts, Evals, and runtime impact.
- Select the corresponding static, structural, evaluation, runtime, and manual review evidence.
- Record the commands executed, results, items not executed, and reasons.
- Output residual risks, as well as the boundaries of conclusions for "can claim/cannot claim".
Core Constraints
skill-up validateis not equivalent to runtime semantic verification.- When project commands are unknown, mark as pending confirmation; do not guess.
- The verification scope must cover the actual changes; do not hide gaps by running full checks.
Load on Demand
Read before producing prompts/skill-change-verification.md; read when evaluation is needed evals/。