The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to manage internal project metadata using the mv command for directories within the .skills/ folder.
[INDIRECT_PROMPT_INJECTION]: The skill processes external information such as frame-change outcomes and roadmap indexes to generate requirements, which creates a potential surface for indirect prompt injection. 1. Ingestion points: Information is ingested from approved frame-change outcomes and docs/roadmap/INDEX.md. 2. Boundary markers: The skill uses markdown headers and a predefined requirements template to structure content. 3. Capability inventory: The agent has the capability to write to the file system and perform basic directory operations. 4. Sanitization: The process includes several mandatory self-review scans (ambiguity, testability, and placeholder checks) and requires an explicit user approval gate before a requirement document is finalized.
[SAFE]: The skill operates entirely within the local project environment using established paths and templates. No remote code execution, unauthorized network access, or credential exposure were identified.

specify-behavior