The Agent Skills Directory

[SAFE]: No security threats were identified. The skill is designed for internal data processing of project management documents (docs/plans/*.md). It does not request network access, execute shell commands, or interact with external APIs.
[DATA_EXPOSURE]: The skill's file access is limited to the docs/plans/ directory for reading and updating workflow plans. This behavior is consistent with its stated purpose and does not involve harvesting credentials, environment variables, or sensitive user data.
[INDIRECT_PROMPT_INJECTION]: While the skill ingests external data from plan files, it implements rigorous validation steps, including cycle detection (topological sorting) and ID verification, to ensure the integrity of the generated dependency graph. It lacks high-risk capabilities (like code execution) that could be exploited via malicious input in the plan documents.

decompose