The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted content from project files (v1 plans) to generate refined task lists. This creates a surface for indirect prompt injection where malicious instructions within the source plan could influence the agent's behavior during the refinement process.
Ingestion points: The skill reads v1 plan files from the project root or docs/ directory using the Read and Glob tools.
Boundary markers: The instructions do not specify any delimiters or safety warnings to treat the plan content as untrusted data.
Capability inventory: The skill has capabilities to write files (Write), read project data (Read, Glob, Grep), and utilize project-specific query tools (gbrain-team__query).
Sanitization: There are no explicit instructions for the agent to sanitize or validate the content of the v1 plan before processing it into implementation steps.

refine-plan