The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content from untrusted external files and interpolates it into prompts for subsequent AI actions, creating a potential injection surface.
Ingestion points: The skill reads CLAUDE.md, Z01_{feature}_research.md, and Z01_CLARIFY_{feature}_research.md from the local file system.
Boundary markers: None. The instructions state to "Include any constraints, patterns, or forbidden approaches from CLAUDE.md here" directly into the prompt for the planning tool without using delimiters or instructions to ignore embedded commands.
Capability inventory: The skill uses TodoWrite for task tracking and invokes the superpowers:writing-plans skill to generate implementation directives and write new files (Z02_*.md).
Sanitization: No sanitization, validation, or escaping of the ingested file content is performed before interpolation.
[COMMAND_EXECUTION]: The skill orchestrates the execution of other tools and skills based on internal logic.
Evidence: Explicitly instructs the agent to invoke the superpowers:writing-plans skill using the Skill tool and provides a custom TodoWrite command to initialize a checklist.

feature-planning