jianying-draft-schema
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill has a theoretical attack surface because it processes untrusted external data such as media metadata and existing video project drafts.
- Ingestion points: External data is introduced into the agent's context through the output of
media probe --jsonand the content of existing drafts duringeditorinspectoperations. - Boundary markers: The skill establishes a logical boundary by requiring that all media facts (durations, IDs, hashes) be sourced from tool-generated output rather than user-provided text, which mitigates simple direct injection attempts.
- Capability inventory: The skill can execute specific local CLI commands (
job run,project verify,media probe) and write project files to isolated output paths. - Sanitization: The skill mandates that all plans be validated against the
jianying-job/v2schema and processed through a domain model, ensuring that untrusted data is converted into a strictly validated structure before any execution occurs.
Audit Metadata