The Agent Skills Directory

[PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by ingesting untrusted content from various project files and passing it to sub-agents via the Task tool without adequate protection.
Ingestion points: Data is read from design/gdd/game-concept.md, design/art/art-bible.md, and .claude/docs/technical-preferences.md.
Boundary markers: The instructions lack delimiters or explicit warnings for agents to ignore instructions embedded within the ingested text.
Capability inventory: The skill uses Write, Edit, and Task tools, enabling it to modify project files and trigger complex logic through sub-agents (art-director, ux-designer, technical-artist).
Sanitization: No filtering, escaping, or validation is performed on the data before it is interpolated into task prompts.

art-bible