omc-plan

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied descriptions and existing plan files, creating a surface for potential injection of malicious instructions.\n
  • Ingestion points: User-provided task descriptions, interview answers, and existing plan files in the .omc/plans/ directory (SKILL.md).\n
  • Boundary markers: The skill labels all generated artifacts as pending approval and enforces a handoff policy requiring explicit user confirmation before execution (SKILL.md).\n
  • Capability inventory: The skill can explore codebases, invoke subagents (planner, architect, critic, analyst), and delegate execution to other modules like team or ralph (SKILL.md).\n
  • Sanitization: A multi-perspective consensus loop and a mandatory structured approval UI are used to validate and approve plans before implementation begins (SKILL.md).\n- [EXTERNAL_DOWNLOADS]: The skill references documentation and changelogs from official domains.\n
  • Evidence: Refers to https://code.claude.com/docs/en/goal and https://raw.githubusercontent.com/anthropics/claude-code/main/CHANGELOG.md for evaluating goal facts (SKILL.md).\n- [COMMAND_EXECUTION]: The skill allows the use of external provider CLIs for subagent roles via specific flags.\n
  • Evidence: Flags like --architect codex trigger the command omc ask codex --agent-prompt architect "..." (SKILL.md).
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 12:35 PM
Security Audit — agent-trust-hub — omc-plan