here
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests file paths and line ranges from a local state file to determine which code to read.\n
- Ingestion points: The agent reads the
reffield from~/.youarehere/ref.json.\n - Boundary markers: Absent; the instructions explicitly tell the agent to use the provided string as-is without manual reconstruction.\n
- Capability inventory: Access to read local files and specific line ranges as specified in the reference file.\n
- Sanitization: Implements version validation checking
extensionVersionagainst the skill's metadata version before processing the data.\n- [COMMAND_EXECUTION]: The skill provides the user with maintenance commands to resolve version mismatches.\n - Evidence: The instructions direct the agent to provide
npx skills updateornpx skills removecommands to the user if the extension and skill versions do not match.
Audit Metadata