The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest context from other local modules, such as profile summaries and task lists. While this theoretically allows for indirect prompt injection from those sources, the risk is mitigated by the skill's core architecture, which forces a mandatory 'Confirmation Stage'. In this stage, the AI must explicitly list its understanding and all inferences for the user to verify one-by-one before any execution begins.
[DYNAMIC_EXECUTION]: The skill provides instructions for the agent to locate and run local companion scripts within the repository's internal directory as a fallback integration method. This dynamic discovery is restricted to the local environment and is intended solely for inter-skill communication, not for executing untrusted external code.

task-clarifier