The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or security vulnerabilities were identified in the skill. The logic follows a transparent, user-approved workflow for project planning.
[DATA_EXFILTRATION]: The skill instructions direct the agent to read local project files, documentation, and commit history. This data access is localized to the environment and is used strictly to provide context-aware design suggestions, with no evidence of external transmission.
[COMMAND_EXECUTION]: The skill is designed to write design documents to the docs/plans/ directory and perform git commit operations. These actions are clearly defined as part of the documentation phase of the brainstorming workflow.
[PROMPT_INJECTION]: The skill contains structural constraints, such as a 'HARD-GATE', which prevent the agent from initiating implementation tasks until the design has been presented to and approved by the user. This serves as a safety boundary for the agent's logic flow.

brainstorming