The Agent Skills Directory

[SAFE]: The skill implements a robust security model based on explicit approval gates. Instructions in SKILL.md and references/risk-gates.md require the agent to seek user confirmation before performing any potentially risky actions, such as modifying the repository, accessing credentials, or executing external mutations.
[COMMAND_EXECUTION]: The skill uses local Python scripts (new_workflow.py, collect_results.py, verify_workflow.py) to manage its workflow artifact system. These scripts are deterministic and operate strictly within the user-authorized project directory.
[SAFE]: The scripts/codex_workflow.py script performs static extraction of workflow metadata from JavaScript-like files. It uses ast.literal_eval for safe data parsing and explicitly rejects executable JS patterns and template interpolation, preventing potential code execution risks from untrusted input files.
[COMMAND_EXECUTION]: The test suite in tests/test_workflow_scripts.py utilizes subprocess.run to verify the functionality of the skill's internal scripts. This usage is confined to the local development environment and is standard practice for verifying tool behavior.

ultracode