improve-codebase-architecture
Pass
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from repository PRs, issues, and comments to generate architectural insights.
- Ingestion points: DEEPENING.md specifies scanning recent PRs and issues for patterns.
- Boundary markers: There are no explicit instructions or delimiters to isolate untrusted content from the agent's internal instructions.
- Capability inventory: The skill instructions in HTML-REPORT.md involve reading codebase history and writing a local ARCHITECTURE_REPORT.html file.
- Sanitization: The skill does not instruct the agent to sanitize or escape user-controlled text (e.g., PR/issue titles) before including them in the HTML report, which could lead to stored XSS when the report is viewed.
Audit Metadata