improve-codebase-architecture
Pass
Audited by Gen Agent Trust Hub on Aug 9, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill instructions emphasize security and read-only access to the codebase. It specifically instructs the agent to treat repository content as evidence to evaluate, not commands to obey, which helps prevent prompt injection from the source code.\n- [SAFE]: Audit results are written to the system's temporary directory by default, ensuring that sensitive reports are not accidentally included in the repository structure.\n- [PROMPT_INJECTION]: The skill identifies and mitigates the risk of indirect prompt injection from codebase data.\n
- Ingestion points: Reads repository files, architecture decisions, and design documents (SKILL.md).\n
- Boundary markers: Explicitly instructs the agent to treat all repository content as evidence to evaluate, not commands to obey (SKILL.md).\n
- Capability inventory: Generates static HTML files and opens them using OS-native commands like open, xdg-open, or start (references/html-report.md).\n
- Sanitization: Mandates escaping all repository-derived strings, implements a restrictive Content Security Policy (CSP), and prohibits remote scripts, fonts, and styles (references/html-report.md).
Audit Metadata