final-cut-pro
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides Python templates and logic for generating FCPXML files based on external inputs like
project_nameandasset_uri. - Ingestion points: User-supplied strings for project names, asset URIs, and marker values in
SKILL.md. - Boundary markers: None present; instructions do not specify delimiters for isolating untrusted metadata.
- Capability inventory: The skill uses
xml.etree.ElementTreeto perform local file-system writes (tree.write) of the generated XML files inSKILL.md. - Sanitization: The Python script uses a standard XML library which provides basic character escaping, though it does not explicitly validate against malicious injection patterns in input strings.
- [SAFE]: The instructions across
SKILL.md,gemini.md, andgpt.mdrepeatedly and explicitly direct the agent to refuse any requests to invent or solicit Apple ID/App Store credentials, mitigating common social engineering vectors in NLE workflows. - [SAFE]: All external documentation references point to trusted official Apple domains (
support.apple.com,developer.apple.com).
Audit Metadata