pr-review-canvas

Pass

Audited by Gen Agent Trust Hub on Jun 21, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local shell commands using gh api to retrieve Pull Request metadata, file diffs, and comments. It also uses jq for data processing and python3 for file assembly and starting a local web server (127.0.0.1:8432). These are standard operations for its intended functionality.\n- [PROMPT_INJECTION]: The skill processes untrusted external content from GitHub (PR titles, descriptions, and comments) which is interpolated into the generated HTML walkthrough. This represents an indirect prompt injection surface.\n
  • Ingestion points: GitHub API responses (metadata and comments).\n
  • Boundary markers: The skill uses a multi-step assembly process that separates data fetching from HTML generation using temporary files.\n
  • Capability inventory: Shell command execution (gh, jq, python3), local file system access (writing to /tmp), and local network binding for the HTTP server.\n
  • Sanitization: The skill explicitly implements JSON sanitization (escaping <, >, and &) in the Python assembly script and HTML character escaping in its JavaScript renderer to prevent cross-site scripting (XSS) or script termination attacks within the generated walkthrough.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 21, 2026, 05:33 PM
Security Audit — agent-trust-hub — pr-review-canvas