acceptance
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes local workspace files such as PROJECT.md, package.json, and various configuration files to determine project categories and recommend testing strategies. While these are local project files, they represent an untrusted data surface that could influence agent behavior. The skill mitigates this by requiring mandatory interactive user confirmation before adopting recommendations or writing files. \n
- Ingestion points: The skill reads workspace configuration and documentation files (e.g., .agents/docs/PROJECT.md, package.json, pyproject.toml, etc.). \n
- Boundary markers: No specific boundary markers are used when reading files, but the skill relies on user confirmation for all actions. \n
- Capability inventory: The skill can write files to the .agents/docs/ directory and recommends shell commands. \n
- Sanitization: The skill uses static templates for script generation rather than directly interpolating raw file content into executable blocks.\n- [COMMAND_EXECUTION]: The skill executes a local script (node .agents/scripts/precheck.mjs) to verify the project state and identifies existing test commands within the workspace for future execution. This is part of the intended project analysis workflow and relies on files expected to be present in the workspace environment.
Audit Metadata