kling-3-0
Pass
Audited by Gen Agent Trust Hub on Aug 13, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill directs the user to install the
@runcomfy/clipackage via npm, which is a tool provided by the vendor, genmedia-labs. - [COMMAND_EXECUTION]: The instructions require the agent to execute shell commands using the
runcomfyCLI to interact with remote video generation services and manage local output files. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted user prompts and external image URLs to drive generative AI outputs.
- Ingestion points: Untrusted data enters the agent context through the
promptandimage_urlfields described in the input schema ofSKILL.md. - Boundary markers: The skill employs JSON-formatted strings via the
--inputflag to separate data from CLI arguments, and the 'Security & Privacy' section explicitly warns that image-based injection is a known risk. - Capability inventory: The skill possesses the capability to execute the
runcomfyCLI, initiate network requests tomodel-api.runcomfy.net, and write data to the local file system via the--output-dirparameter. - Sanitization: The documentation notes that the CLI avoids shell expansion for input parameters, providing basic protection against command injection, though it lacks specific sanitization for the content of processed images.
Audit Metadata