The Agent Skills Directory

[SAFE]: The skill operates as a document generator, creating markdown-based testing protocols for manual execution by humans.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill reads project context from files like technical preferences and design documents to tailor the protocol. It does not access sensitive system paths, environment variables containing secrets, or hardcoded credentials.
[INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect injection as it processes project documentation and playtest logs.
Ingestion points: .claude/docs/technical-preferences.md, design/gdd/game-concept.md, and files in production/playtests/ and production/qa/.
Boundary markers: Absent.
Capability inventory: Read, Glob, Grep, and Write tools.
Sanitization: Absent.
Assessment: The risk is minimal because the generated output is a static markdown file intended for human review and manual data entry, providing no path for automated exploitation.

soak-test