skill-official-evaluation
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because its primary function is to ingest and analyze untrusted content from external skill directories provided by the user.
- Ingestion points: The
scripts/official_evaluate.pyscript reads the content ofSKILL.mdand other text-based files (scripts, references) from a user-specified path via the--skill-dirargument. - Boundary markers: The skill instructions and evaluation script do not implement explicit boundary markers or 'ignore' instructions to isolate the untrusted data being processed from the agent's core logic.
- Capability inventory: The skill possesses the capability to read the local filesystem and write output files or report findings to the agent's context.
- Sanitization: The script truncates long strings for evidence presentation in the report but lacks robust sanitization to prevent instructions inside the analyzed skill from influencing the agent's evaluation process.
Audit Metadata