mmx-h3-video
Pass
Audited by Gen Agent Trust Hub on Aug 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill interacts with the MiniMax-H3 API through shell commands using the
mmxCLI. It provides specific instructions for building a local artifact (bun run build) and executing the resulting JavaScript file (node ./dist/mmx.mjs) to perform generation and configuration tasks. - [PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection through its command construction methods.
- Ingestion points: User-supplied text is interpolated into the
--promptargument of shell commands described in SKILL.md. - Boundary markers: The skill's examples utilize double quotes to encapsulate the prompt (e.g.,
--prompt "<video prompt>"), which does not protect against shell metacharacters such as backticks or subshells. - Capability inventory: The agent is authorized to execute shell commands via the
mmxandnoderuntimes. - Sanitization: There are no explicit instructions within the skill to sanitize, escape, or validate user-provided strings for shell-active characters prior to execution.
Audit Metadata