The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it parses content from various user-provided markdown files to drive its logic and sub-agent instructions.
Ingestion points: The skill reads project-local documentation from paths including design/gdd/, design/art/, design/levels/, and design/narrative/ (found in references/claude-skill.md).
Boundary markers: There are no explicit delimiters or instructions provided to the agent to disregard embedded commands or formatting within the source documentation.
Capability inventory: The skill has the ability to write/edit project files and uses the Task tool to spawn sub-agents (art-director, technical-artist) with instructions derived from the ingested data.
Sanitization: No content validation or sanitization is performed on the ingested text before it is used to generate prompts for the sub-agents.

asset-spec