prompt-optimizer
Pass
Audited by Gen Agent Trust Hub on Jun 24, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill reads untrusted project metadata files (e.g.,
CLAUDE.md, dependency manifests) to extract context for tech stack detection. While these files could contain malicious instructions if the user is working within a compromised repository, the skill's strict advisory-only constraints and predefined output structure mitigate the risk of instruction override. - [COMMAND_EXECUTION]: The skill incorporates specific tool command recommendations (e.g.,
/plan,/tdd,/verify) into its output. These are presented as part of an optimized prompt for the user's reference and are not invoked by the skill itself during execution.
Audit Metadata