The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill utilizes the !bash syntax in SKILL.md to execute a shell script named yaml-helper.sh at load time. This command is used to resolve configuration keys such as review_mode and workflow. While the command appears to be a legitimate part of the project's configuration management, load-time execution represents an attack surface for silent command execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data that could contain malicious instructions.
Ingestion points: The skill reads design/gdd/game-concept.md (or design/game-brief.md) in Phase 0 and Phase 5.
Boundary markers: No explicit delimiters or instructions are used to warn the agent to ignore embedded commands within the ingested content when it is passed to sub-agents like the art-director.
Capability inventory: The skill and its sub-agents have access to powerful tools including Write, Edit, Bash, and the ability to spawn further agents.
Sanitization: There is no evidence of sanitization or validation of the content extracted from the game concept files before it is interpolated into prompts.

art-bible