The Agent Skills Directory

[SAFE]: The skill implements a structured validation workflow using human-in-the-loop gates and evidence-based reasoning. Analysis of all skill files shows no evidence of malicious intent or safety violations.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided requirements and external specification files. This risk is effectively mitigated by Step 4, which requires a blocking human confirmation gate, and a mandatory requirement for all claims to cite specific evidence with high confidence scores.
[DYNAMIC_EXECUTION]: The skill dynamically generates prompts for sub-agents using predefined templates for specialized reviews. This execution is confined to internal platform tools and follows intended architectural patterns for multi-agent systems.

spec-clarify