agentcanon-manifest
Pass
Audited by Gen Agent Trust Hub on Aug 19, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it crawls the local filesystem and ingests data from external SKILL.md files.
- Ingestion points: The agent reads the name and description fields from every SKILL.md file located in ~/.agents/skills/ and other project directories.
- Boundary markers: The instructions do not specify any delimiters or safety warnings to ignore instructions embedded in the scanned files.
- Capability inventory: The skill possesses read access to the home directory and write access to the ~/.agents/ directory.
- Sanitization: No explicit sanitization or filtering is applied to the content extracted from the external files before it is incorporated into the manifest.
- [DATA_EXPOSURE]: The skill performs a wide-ranging scan of the user's home directory to locate skill folders, extracting absolute file paths and filesystem timestamps to build the manifest record.
Audit Metadata