r3f-game-director
Pass
Audited by Gen Agent Trust Hub on Jun 26, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill consists entirely of markdown-based instructions for project orchestration and does not include any executable scripts, binaries, or remote dependencies.
- [PROMPT_INJECTION]: The skill establishes an orchestration workflow that ingests content from sibling skill files, which creates a surface for indirect prompt injection. \n
- Ingestion points: Sibling
SKILL.mdand reference files are loaded from the parent directory or~/.claude/skills/. \n - Boundary markers: None identified; instructions from loaded sibling skills are integrated directly into the current execution context without delimiters. \n
- Capability inventory: The orchestrated agent is authorized to modify local source code (e.g.,
Scene.tsx) and delegate tasks to subagents. \n - Sanitization: No validation or sanitization of the content from loaded sibling skills is performed.
Audit Metadata