skills/jscraik/agent-skills/autofix/Gen Agent Trust Hub

autofix

Pass

Audited by Gen Agent Trust Hub on Apr 12, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill utilizes the official GitHub CLI (gh) for platform interactions. The associated helper script, scripts/fetch_unresolved_threads.py, uses subprocess.run with argument lists and no shell environment, which is a secure practice that prevents command injection.
  • [PROMPT_INJECTION]: The skill is exposed to untrusted data from review comments. Evidence chain: (1) Ingestion point: scripts/fetch_unresolved_threads.py extracts comment bodies from GitHub. (2) Boundary markers: Not explicitly defined in prompt interpolation. (3) Capability inventory: Uses subprocess calls, git operations, and file writing. (4) Sanitization: Constraints require redaction of secrets. The risk of indirect prompt injection is successfully mitigated by a mandatory human-in-the-loop approval process and explicit instructions to disregard arbitrary commands within comments.
  • [CREDENTIALS_UNSAFE]: No hardcoded secrets or unauthorized access to sensitive file paths (e.g., .ssh, .aws) were detected. The skill specifically instructs the agent to redact credentials from all outputs.
  • [EXTERNAL_DOWNLOADS]: The skill does not perform any unverified remote downloads. It relies on a local Python script and the standard GitHub CLI tool.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 12, 2026, 01:13 PM
Security Audit — agent-trust-hub — autofix