game-character-pipeline

Pass

Audited by Gen Agent Trust Hub on Nov 20, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a coordination workflow that consumes external, potentially untrusted project data (such as ART_DIRECTION.md, project briefs, and ledger notes) to guide the character generation and rigging pipeline. Instructions embedded in these project files could influence the agent's decisions during stage transitions or gate approvals without sufficient boundaries.
  • Ingestion points: SKILL.md and reference/start-gate.md indicate the agent reads core instructions from ART_DIRECTION.md, user-provided briefs, and the ledger's note field.
  • Boundary markers: The instructions do not define specific delimiters or "ignore embedded instructions" warnings for the agent when processing these project-level text files.
  • Capability inventory: The documented pipeline triggers the execution of numerous Python scripts (e.g., autorig.py, twocam.py, mesh_check.py, contract_map.py) and interacts with external tools like Blender, Unity, and ComfyUI.
  • Sanitization: There is no evidence of input validation or content sanitization for the project instructions used to configure the pipeline backend or quality targets.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 20, 2026, 03:25 PM