middle
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection because it is designed to ingest, analyze, and modify untrusted source code from a user's project.\n
- Ingestion points: Reads user-specified project files and configuration manifests as defined in the SKILL.md Operating Rules section 2.\n
- Boundary markers: There are no explicit instructions to delimit external content or to provide guardrails against instructions embedded within the analyzed data.\n
- Capability inventory: The skill is empowered to modify existing project files and execute local validation tools such as linters, typecheckers, and build scripts upon approval.\n
- Sanitization: The instructions do not define sanitization or escaping protocols for ingested content before it enters the agent's context.\n- [COMMAND_EXECUTION]: The skill instructions define a behavior where the agent performs file system writes and executes shell commands for code validation (e.g., build, typecheck, lint) on the user's system.
Audit Metadata