repository-readme-architect-acrazie
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from the repository to inform documentation generation.
- Ingestion points: The agent reads repository manifests, package metadata, scripts, and code files from the local filesystem.
- Boundary markers: The skill defines a 'Project loading boundary' to prevent unauthorized instruction execution and uses explicit approval gates for user decisions.
- Capability inventory: Capabilities are restricted to read-only inspection and writing specifically to the primary README.md file; no other files can be modified.
- Sanitization: Instructions require validating Markdown structure and anchors, and explicitly forbid the exposure of secrets found during inspection.
- [COMMAND_EXECUTION]: The skill allows for limited command execution to verify documentation instructions.
- Evidence: SKILL.md specifies running only 'safe, cheap commands' and prohibits network access, dependency installation, or mutating operations without further permission.
- [DATA_EXPOSURE]: The skill instructions involve reading repository configuration files.
- Evidence: SKILL.md contains strict requirements to never expose secrets, avoid secret-bearing files where possible, and use only dummy data in generated examples.
Audit Metadata