agent-operating-skill
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill provides guidelines for using system tools (read_file, write_file, run_command) to complete repository-based tasks. It mandates adherence to safety refusals and permission boundaries defined by the environment.\n- [DYNAMIC_EXECUTION]: The instructions support delegating tasks to sub-agents via tools like workflow.py dispatch and codex exec. This is managed through strictly defined handoff capsules and ownership contracts to prevent unauthorized state changes.\n- [INDIRECT_PROMPT_INJECTION]: The skill manages a mailbox for cross-runtime communication. It mitigates injection risks by instructing agents to treat incoming messages as context only and to explicitly verify they fit within the current user-approved scope before execution.\n- [SAFE]: The skill contains specific directives to protect sensitive data, including secrets and credentials, and emphasizes following the project's existing security and architecture rules.
Audit Metadata