codegraph-ast-grep

Pass

Audited by Gen Agent Trust Hub on Nov 9, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill facilitates the download of CodeGraph and ast-grep from official GitHub releases and package registries (NPM, PyPI, Cargo). It strictly enforces a verification-first approach, requiring SHA-256 checksum validation and manual inspection of archives before extraction.
  • [COMMAND_EXECUTION]: The agent is instructed to execute shell commands for installation and project initialization. Safety is maintained through idempotent workflows, the use of project-local shims to avoid path shadowing, and the requirement for explicit user approval for any mutation or privilege escalation.
  • [DATA_EXFILTRATION]: The skill explicitly addresses data privacy by defaulting to CODEGRAPH_TELEMETRY=0 and instructing the agent to redact secrets, internal hostnames, and customer data from any persisted artifacts or public logs.
  • [INDIRECT_PROMPT_INJECTION]: While the skill ingests untrusted repository content for analysis, it defines clear capability boundaries, requires user approval for all write operations, and creates persistent guidance to help downstream agents interpret results safely.
  • [SAFE]: The skill demonstrates a mature security posture by rejecting 'pipe-to-shell' installation patterns and providing detailed troubleshooting steps that prioritize non-destructive diagnostics.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 9, 2026, 07:19 AM