faceless-explainer
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill orchestrates a multi-stage video creation process using local Node.js scripts and invokes standard utilities like
ffmpegandffprobefor media handling, which is consistent with its primary function.\n- [SAFE]: External library dependencies are restricted to the GSAP library loaded via the well-known and trusted jsDelivr CDN, adhering to security best practices for web-based resources.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests arbitrary text inputs from users to serve as source material for video scripts. This constitutes a potential indirect prompt injection surface; however, the risk is assessed as safe because the workflow necessitates significant restructuring and summarization by the agent before the content is used in any visual or audio generation step.\n - Ingestion points: User-supplied text is stored in
capture/extracted/visible-text.txt(SKILL.md).\n - Boundary markers: The workflow relies on natural language boundaries and agent interpretation rather than explicit technical delimiters.\n
- Capability inventory: The skill has file system access and shell execution capabilities (via
spawnSyncfor media tools) but these are gated by the structured workflow.\n - Sanitization: The agent is explicitly instructed to extract 'teaching truth' and restructure the source text, which effectively buffers the system from malicious content in the raw input.
Audit Metadata