resolve-fixed-pr-comments

Pass

Audited by Gen Agent Trust Hub on Jun 21, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses gh CLI and git commands to perform read-only operations such as viewing PR status and inspecting file content.
  • [EXTERNAL_DOWNLOADS]: Includes a fallback mechanism using curl to fetch information from public GitHub repositories when CLI or MCP tools are unavailable.
  • [PROMPT_INJECTION]: The skill processes untrusted PR comments, which creates a potential surface for indirect prompt injection.
  • Ingestion points: PR review comments and thread descriptions are ingested from GitHub (SKILL.md).
  • Boundary markers: The instructions include strict guidelines that define the only permitted write action as resolving a thread, effectively instructing the agent to ignore any executable instructions within the comments (SKILL.md).
  • Capability inventory: The skill is strictly limited to resolving PR threads; it is expressly forbidden from modifying code, creating files, or performing mutating git operations like push or commit (SKILL.md).
  • Sanitization: Verification is based on matching the current code state to human/bot requirements rather than executing or interpolating comment text as commands.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 21, 2026, 12:11 AM
Security Audit — agent-trust-hub — resolve-fixed-pr-comments