jianying-transitions

Pass

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill is built to interact with a specific local command-line interface tool (jianying). It implements security gates by requiring version checks, capability discovery (supported status), and explicit user approval before executing any command that has side effects or costs.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as video project files, media metadata, and transition catalogues.
  • Ingestion points: Data is ingested via media catalog queries and project file probing (SKILL.md, references/workflow.md).
  • Boundary markers: The skill explicitly instructs the agent to validate schemas and reject unknown or out-of-bounds input (Step 3).
  • Capability inventory: Capabilities include file reads/writes (restricted to isolated copies) and CLI execution.
  • Sanitization: Implements strict validation of media facts, hashes, and project versions before processing, which significantly mitigates risks associated with malicious project data.
  • [SAFE]: The skill explicitly forbids dangerous behaviors such as generating dynamic Python scripts, calling headless browsers, or overriding original source files without authorization.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 19, 2026, 11:55 PM
Security Audit — agent-trust-hub — jianying-transitions