temps-mcp-setup
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides legitimate instructions for using
bunxto run the@temps-sdk/clitool, which is necessary for managing the Temps MCP server and client configurations. - [EXTERNAL_DOWNLOADS]: The skill utilizes the
@temps-sdk/clipackage viabunx, which is a standard method for accessing the official project utility. This download is directly related to the skill's primary purpose of setting up the Temps environment. - [INDIRECT_PROMPT_INJECTION]: The skill enables an indirect prompt injection surface by adding MCP tools to the agent context. However, this is managed safely through several mitigations:
- Ingestion points: The agent processes project and deployment data retrieved from the Temps API via MCP tools.
- Boundary markers: All write-based tool calls (e.g., triggering deployments) require explicit human confirmation via a time-limited token, preventing unauthorized autonomous actions.
- Capability inventory: The tools are limited to the specific scopes defined by the user-provided API key (e.g.,
DeploymentsWrite). - Sanitization: Authentication and authorization are handled by the server-side Temps service layer, ensuring the agent cannot exceed the key's granted permissions.
Audit Metadata