muapi-product-campaign
Warn
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: MEDIUMINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it incorporates raw user inputs into instructions sent to sub-models.
- Ingestion points: product_name, campaign_message, target_audience, and visual_style inputs in SKILL.md.
- Boundary markers: Absent; inputs are directly concatenated into prompts.
- Capability inventory: The skill executes muapi CLI commands and curl requests as defined in SKILL.md.
- Sanitization: None; there are no instructions to filter or escape user content.
- [COMMAND_EXECUTION]: The skill's design pattern for calling the muapi CLI and curl involves direct string substitution of user-provided variables into a shell command line. This creates a significant surface for command injection attacks if the agent does not correctly sanitize the inputs.
- [DYNAMIC_EXECUTION]: Instructions in the 'Notes for the Executing Agent' section specify that the agent should manually substitute placeholders into shell commands at runtime, which is a form of dynamic command assembly using untrusted data.
Audit Metadata