The Agent Skills Directory

[SAFE]: The skill defines a markdown-based planning methodology using files like task_plan.md and notes.md. This is a productivity-enhancing design pattern aimed at improving context retention for AI agents and does not involve harmful logic.
[COMMAND_EXECUTION]: Several files contain markdown code blocks labeled as bash (e.g., Read task_plan.md, Edit task_plan.md). These are descriptive pseudocode examples intended to demonstrate agent actions/tool-calls rather than executable shell scripts. No actual shell commands or command injection paths were found.
[DATA_EXPOSURE]: Analysis of the skill body and metadata found no hardcoded credentials, sensitive file path access (e.g., .ssh or .env), or unauthorized network operations.
[PROMPT_INJECTION]: The skill contains instructional rules (e.g., 'ALWAYS Create Plan First') which are operational guidelines for the workflow. There are no attempts to override system safety filters or disregard prior instructions.
[INDIRECT_PROMPT_INJECTION]: The skill involves reading external research findings from notes.md. While this ingestion of external data constitutes an attack surface, the risk is categorized as low because the skill is focused on task organization and does not include high-privilege capabilities or evidence of malicious exploitation of that data.

planning-with-files