agent-capability-discovery
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest metadata from arbitrary
skill.yamlfiles, which serves as a vector for indirect prompt injection. - Ingestion points: Files are read from the
base_directoryinput, specifically targeted atskill.yamldefinitions across the local repository. - Boundary markers: The output format (markdown, JSON, or YAML) does not implement explicit delimiters or "ignore embedded instructions" warnings to prevent the agent from treating discovered metadata as authoritative commands.
- Capability inventory: The skill performs recursive directory traversal and file reading to extract descriptive text.
- Sanitization: There is no evidence of sanitization for the 'capabilities' or 'description' fields extracted from external files before they are presented to the master agent for decision-making.
Audit Metadata