autopilot-tickets

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external ticket trackers (comments, acceptance criteria) and incorporates this information into its decision-making loop.\n
  • Ingestion points: SKILL.md (Step 2: "Query open 'ready-for-agent' tickets... Read their acceptance criteria, labels, recent comments...")\n
  • Capability inventory: Git repository management (git status, git worktree), pull request promotion and merging via forge tools, and subagent lifecycle management.\n
  • Boundary markers: No specific delimiters or "ignore instructions" directives are used when processing external ticket content.\n
  • Sanitization: No evidence of sanitization or validation of the ingested ticket data before it influences the agent's logic.\n- [COMMAND_EXECUTION]: The skill explicitly instructs the agent to execute shell commands for repository state management and cleanup.\n
  • Evidence: git status --short --branch and git worktree remove are specified in the loop instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 01:01 AM
Security Audit — agent-trust-hub — autopilot-tickets