three-fiber-component-shape-planner
Pass
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: No patterns of instruction overriding, safety bypass, or system prompt extraction were detected. The skill maintains strict boundaries and explicitly refuses to generate implementation code when requested.
- [DATA_EXFILTRATION]: No sensitive file paths, environment variable access, or hardcoded credentials were found. There are no network operations (curl, wget, etc.) present in the instructions.
- [OBFUSCATION]: The content is clear and uses standard Markdown formatting. No Base64 encoding, zero-width characters, or homoglyph-based URL spoofing were detected.
- [REMOTE_CODE_EXECUTION]: The skill is designed to produce planning documentation ('specs') only. It contains explicit prohibitions against generating component implementation code and does not invoke any package managers or remote scripts.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process data from other internal handoff skills (animation contracts, performance budgets).
- Ingestion points: Data enters via upstream contract files (e.g., glb-animation-handoff).
- Boundary markers: None explicitly defined, though instructions treat input as data to be tabulated.
- Capability inventory: No dangerous capabilities (file writes, network, shell execution) are used; the skill only outputs text.
- Sanitization: None detected, but the lack of executable capabilities renders the risk negligible.
Audit Metadata