muapi-workflow
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill operates by wrapping the muapi CLI tool within several shell scripts, such as generate-workflow.sh and run-workflow.sh. This is the intended behavior for providing an interface to the workflow engine.
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it processes natural language descriptions to create and modify workflows.
- Ingestion points: Natural language prompts passed to muapi workflow create and muapi workflow edit in SKILL.md and generate-workflow.sh.
- Boundary markers: The SKILL.md file contains a 'CRITICAL RULE' instructing the agent to confirm input values with the user and never guess them, which mitigates the risk of the agent autonomously executing injected commands.
- Capability inventory: The skill can execute commands via the muapi CLI, write files to the local system using the --download option, and make outbound network requests using the --webhook parameter.
- Sanitization: No specific prompt sanitization or escaping mechanisms are implemented in the shell script wrappers.
- [DATA_EXFILTRATION]: The run-workflow.sh script facilitates potential data exfiltration by supporting a --webhook parameter, which can send workflow results to an external URL, and a --download parameter for local file storage.
Audit Metadata