The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external data, which introduces a potential surface for indirect prompt injection.
Ingestion points: The skill performs codebase research and requirement clarification in Phase 0 and Phase 1 (SKILL.md).
Boundary markers: The 'Safety Boundaries' section provides instructions to prevent the agent from inventing data or overstepping its planning role.
Capability inventory: The skill frontmatter defines a file_write permission, allowing the agent to persist artifacts to the file system.
Sanitization: The instructions lack specific guidance on escaping or validating codebase content before it is processed or reflected in generated plans.

planner