The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill conducts an interview and writes user-provided text directly into the AGENTS.md file, which serves as a global instruction set for the agent in the environment. This creates a surface where future agent behavior is governed by data provided during the setup phase.
Ingestion points: User input from the 5-question interview in SKILL.md (Step 2).
Boundary markers: None; user input is interpolated directly into markdown headers and bullet points.
Capability inventory: Local file system writes, directory creation, and git repository management.
Sanitization: No escaping or validation is performed on the user-supplied strings before writing them to the configuration files.
[DYNAMIC_EXECUTION]: The skill programmatically generates a secondary agent skill (.agents/skills/projects/SKILL.md) and configures the environment to load it. This is a mechanism for the agent to extend its own functional capabilities based on a static template.
Evidence: Step 6 generates a new SKILL.md with its own logic for managing project lifecycles.

setup-brain