blender-composition-camera-planner
Pass
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user data in the form of scene briefs and subject intent to generate composition plans. While it lacks explicit boundary markers or input sanitization, the risk is mitigated by the skill's design as a planning-only tool with no access to code execution, shell commands, or network operations.
- Ingestion points: SKILL.md (Required inputs: subject intent, target output context)
- Boundary markers: Absent
- Capability inventory: Planning only; no subprocess calls, exec/eval, file-write, or network capabilities are permitted per the skill's constraints.
- Sanitization: Absent
- [NO_CODE]: The skill consists entirely of instructional markdown and reference files. It contains explicit prohibitions against writing Python code or Blender execution commands in its anti-patterns and non-goals sections, significantly reducing the attack surface.
- [SAFE]: All external references target official documentation and well-known services (Blender.org, Claude.com). No obfuscation, persistence mechanisms, or credential exposure patterns were detected.
Audit Metadata