gh-address-comments

Pass

Audited by Gen Agent Trust Hub on Apr 5, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted external data (GitHub comments) which can lead to indirect prompt injection.
  • Ingestion points: Pull request comments, reviews, and threads are fetched from GitHub via scripts/fetch_comments.py and provided to the agent context.
  • Boundary markers: Absent. The instructions do not provide delimiters or warnings to help the agent distinguish between legitimate task descriptions and malicious instructions embedded within the fetched comments.
  • Capability inventory: The agent is empowered to "Apply fixes" (file system modification) based on the fetched data and executes shell commands via the gh CLI (scripts/fetch_comments.py).
  • Sanitization: Absent. The skill passes fetched content directly to the agent without filtering or validation.
  • [COMMAND_EXECUTION]: The script scripts/fetch_comments.py utilizes subprocess.run() to execute gh (GitHub CLI) commands. While it avoids shell=True and passes arguments as a list to prevent basic command injection, it relies on the execution of external CLI tools with the user's local authentication state.
  • [PRIVILEGE_ESCALATION]: The SKILL.md file explicitly requests escalated platform permissions (sandbox_permissions=require_escalated) and elevated network access to enable the gh CLI to operate outside the default restricted sandbox environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 5, 2026, 05:14 PM
Security Audit — agent-trust-hub — gh-address-comments