scenario-godot-vfx
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external Godot scene files (.tscn) and node hierarchies to perform audits and automated rendering. This creates a surface for indirect prompt injection if a scene contains malicious metadata or node configurations designed to influence the agent.
- Ingestion points: Scene paths and node names passed to audit, timeline, fit_aabb, and stress jobs in scripts/agentkit/vfx/vfx_jobs.gd and scripts/agentkit/vfx/vfx_capture.gd.
- Boundary markers: None. The skill assumes the provided scene files are trusted Godot assets.
- Capability inventory: The skill has the ability to write Godot scenes to disk (via AgentBuild.save_scene) and execute Godot scripts (via the gd_run toolkit).
- Sanitization: There is no sanitization of scene content or node properties before they are processed by the audit and capture scripts.
- [DYNAMIC_EXECUTION]: The fit_aabb and timeline methods in scripts/agentkit/vfx/vfx_capture.gd use Object.call() to dynamically invoke methods on Godot nodes. The method names are supplied via job arguments (play_method), allowing for dynamic code execution within the Godot runtime environment.
- [EXTERNAL_DOWNLOADS]: The SKILL.md file suggests using npx skills add scenario-labs/skills to install missing sibling skills. These references point to the vendor's own established infrastructure and are used for maintaining the toolkit's dependencies.
Audit Metadata