vision
Pass
Audited by Gen Agent Trust Hub on Nov 12, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to run a local linter script using
python3 docs/ux/lint.py. The instructions specify that this script is seeded by the framework, uses only the Python standard library, and does not require external package installation.\n- [INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface by reading project documentation, source code, and configuration files to derive a product vision. \n - Ingestion points: Project files including
README.md,ARCHITECTURE.md,ROADMAP.md,CHANGELOG.md,package.json,pyproject.toml, and source code entry points.\n - Boundary markers: None identified for the reading phase.\n
- Capability inventory: The skill has file read/write permissions and modifies agent instruction files (
CLAUDE.md,AGENTS.md,GEMINI.md).\n - Sanitization: None identified; the skill relies on the agent to interpret the data correctly for UX purposes.\n- [PERSISTENCE]: The skill installs a persistent "vision-alignment rule" into the agent's primary instruction files (e.g.,
CLAUDE.md,AGENTS.md,GEMINI.md). This is a legitimate functional feature designed to ensure the agent evaluates future project changes against the established product vision document.
Audit Metadata