code-map

Pass

Audited by Gen Agent Trust Hub on Nov 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: Uses git and gh (GitHub CLI) to retrieve repository state and PR metadata. These calls use Node's execFile for secure subprocess management.
  • [EXTERNAL_DOWNLOADS]: Automates the installation of Node.js libraries from the NPM registry and browser binaries via Playwright.
  • [INDIRECT_PROMPT_INJECTION]: Maintains an attack surface for indirect injection by ingesting untrusted code and PR data from files and the GitHub API. Ingestion points occur in core.mjs via captureLocal and capturePr. Boundary markers are provided as agent instructions in runtime/guide.md. Capability inventory includes shell command execution for analysis. Sanitization is handled by a redact function that removes credentials from captured source code.
  • [DYNAMIC_EXECUTION]: Executes JavaScript within a headless browser to generate Mermaid diagrams from extracted code structure.
  • [SAFE]: Implements several security guardrails, including a checkedPath utility to prevent directory traversal and a robust redaction engine for sensitive metadata like API keys and private keys.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 18, 2026, 03:01 AM