The Agent Skills Directory

[SAFE]: The skill instructions are entirely focused on functional project planning and do not contain any obfuscation, hardcoded credentials, or malicious overrides.
[PROMPT_INJECTION]: The skill is designed to analyze project state by reading existing files and commits, which creates a potential surface for indirect prompt injection from repository content. 1. Ingestion points: Project files, documentation, and recent git commits are read to build context. 2. Boundary markers: The skill instructions enforce an incremental process, breaking designs into small parts for separate validation and requiring user confirmation at each step. 3. Capability inventory: The agent is authorized to write design documents to the file system and perform git operations. 4. Sanitization: There is no programmatic sanitization of project data, but the risk is managed through mandatory human review of the agent's proposed designs.
[COMMAND_EXECUTION]: The skill utilizes git commands to commit documentation and manage worktrees. These operations are standard for development assistants and are used appropriately for persistence and isolation of work.

brainstorming