The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill directs the use of specific Node.js packages including vite-plugin-uni-cloudfunctions for packaging and @cloudbase/cloudbase-mcp for CloudBase integration. These dependencies are standard for the skill's stated development purpose.
[COMMAND_EXECUTION]: The skill instructs the agent to perform validation using development tools such as pnpm, tsc, and vitest. These commands are used to ensure the generated code is correct and follows project conventions.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided business requirements and data schemas to generate code. While this ingestion surface exists, the risk is mitigated by the use of strict architectural patterns and the enforcement of Zod schema validation in the generated functions.
Ingestion points: User-provided business requirements and CloudBase field definitions processed via the SKILL.md contract.
Boundary markers: Absent in prompt interpolation instructions, but mitigated by strict template enforcement.
Capability inventory: File system writes for boilerplate generation and command execution for build/test cycles.
Sanitization: Enforced use of Zod for input validation within the generated cloud function code.
[SAFE]: The skill explicitly handles credentials securely in its mcp-setup.md reference by using placeholders and instructing the agent never to guess or fill in real secrets, ensuring that sensitive CloudBase API keys are managed solely by the end-user.

wx-cloudfunctions