pr-review-canvas
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external GitHub Pull Request data including titles, descriptions, and comments which are user-controlled.
- Ingestion points: Fetches PR metadata and discussion content via gh api calls in SKILL.md.
- Boundary markers: The skill instructions specify a Python assembly step using json.dumps and character replacement for script-sensitive characters to prevent payload breakout.
- Capability inventory: Limited to generating local HTML files and serving them on localhost; no automated execution of external data.
- Sanitization: Implements HTML escaping via a provided esc() utility in renderer.js and script-safe JSON embedding logic.
- [EXTERNAL_DOWNLOADS]: The skill references styles from fonts.googleapis.com in template.html, which is a well-known and trusted service used for the report's UI.
- [COMMAND_EXECUTION]: Utilizes the GitHub CLI (gh) for data retrieval and Python (python3) for local file assembly and hosting. These operations are restricted to the local environment and whitelisted GitHub domains, following standard developer workflows.
Audit Metadata