The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash and Task tools to execute generated prototype code and manage the creative-director review task. These operations are intended for local development within the project worktree.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface as it ingests data from local project files and subsequently executes commands based on that context. This is a standard risk for development-focused skills.
Ingestion points: production/review-mode.txt, CLAUDE.md, .claude/docs/director-gates.md, design/gdd/game-concept.md (via Read/Glob/Grep tools).
Boundary markers: Absent.
Capability inventory: Bash, Task, Write, Edit, Read, Glob, Grep tools.
Sanitization: Absent.

prototype