add-skill
Pass
Audited by Gen Agent Trust Hub on Jul 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The workflow involves running
make checkto validate new skills. This executes local validation logic which is expected for a developer-oriented tool. - [PROMPT_INJECTION]: The skill ingests data from local directories (
.local/), user arguments, and web search results to generate new instructions. This pattern creates an indirect prompt injection surface where untrusted external data could influence the behavior of authored skills. (1) Ingestion points:.local/directory, web search results, and user-provided arguments in SKILL.md. (2) Boundary markers: None explicitly defined for isolating external research data. (3) Capability inventory: Local command execution (make check) and file system write access for skill generation. (4) Sanitization: No explicit sanitization or filtering logic is described for the research phase.
Audit Metadata