jianying-draft-media
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes specialized CLI tools (such as
jianying,project, andtemplate) to perform media imports, replacements, and project modifications. These operations are restricted to fixed arguments and specific semantics defined within the workflow to prevent arbitrary shell injection. - [EXTERNAL_DOWNLOADS]: The skill provides mechanisms to fetch official JianYing resources. These downloads are not arbitrary; they are managed through adapters that require verification of usage rights and documentation of the resource's identity before integration.
- [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect injection as it processes external media assets and project files. However, it implements a robust 'media preflight' protocol including: 1. Ingestion points: External media files (e.g., MP4, WebM) and JianYing project drafts (specified in SKILL.md and workflow.md). 2. Boundary markers: Use of isolated project copies ('project.edit_isolated') and explicit approval status requirements ('approval.status'). 3. Capability inventory: Resource probing and project editing capabilities (e.g.,
media.probe,project.verify) used throughout the command list. 4. Sanitization: Mandatory SHA-256 content hashing to ensure material identity and quality comparisons to detect unexpected changes.
Audit Metadata