The Agent Skills Directory

[SAFE]: The skill is primarily a design and advisory tool with a 'disable-model-invocation: true' configuration, which limits its ability to execute external tools without direct instructions. It focuses on the creation of Markdown artifacts and payloads for other agents to execute, rather than performing high-risk actions itself. Nested safety mechanisms include:
Mandatory user approval ('ok to emit') before any files are written or payloads generated.
Explicit human-in-the-loop requirements for 'Merge/deploy/system' actions.
Use of boundary markers (e.g., ) for critical limits like budgets and iteration caps, ensuring the user provides specific constraints.
A 'Design only' contract that strictly prohibits the skill from starting or executing the designed workflows.

looper