fix
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill reads project files (HTML, JSON-LD, robots.txt) to suggest and apply modifications, creating a potential surface where malicious content within those files could influence the agent's output.
- Ingestion points: Local project files read via the Read and Grep tools.
- Boundary markers: Not explicitly defined for the prompt interpolation logic, though the skill utilizes a separate subagent for writing.
- Capability inventory: Includes Bash, Task, and file-writing capabilities via a dedicated subagent.
- Sanitization: Includes a post-write verification step that runs assertion scripts to validate the integrity of applied changes.
- [COMMAND_EXECUTION]: The skill utilizes Bash and Task tools for administrative operations including file backups, git status monitoring, and executing validation scripts like validate-jsonld.mjs. These capabilities are constrained by safety rules that prevent access to sensitive directories like .git or environment configuration files.
Audit Metadata