z-api
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the user or agent to install the Membrane CLI from the NPM registry and utilizes npx for remote execution of vendor-provided tools.
- Evidence:
npm install -g @membranehq/cli@latestand usage ofnpx @membranehq/cli@latestin SKILL.md. - Context: The package @membranehq/cli is an official resource from the skill author (membranedev).
- [COMMAND_EXECUTION]: The skill relies on executing shell commands to perform all its operations, including authentication, connection management, and data retrieval.
- Evidence: Commands such as
membrane login,membrane connection ensure,membrane action list, andmembrane requestare used throughout SKILL.md. - [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes and interprets data retrieved from external Z-API endpoints.
- Ingestion points: Data returned from
membrane action runandmembrane requestcommands. - Boundary markers: The skill encourages the use of the
--jsonflag to maintain data structure, though it does not explicitly provide LLM-specific delimiters. - Capability inventory: The agent can execute shell commands, read/write files via the CLI, and make network requests via the proxy.
- Sanitization: No explicit sanitization logic is present in the markdown; it relies on standard JSON processing by the agent.
Audit Metadata