The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of standard development dependencies, specifically the official '@vapi-ai/server-sdk' for interacting with the Vapi platform, along with 'typescript' and '@types/bun'. These are well-known packages from legitimate registries.
[COMMAND_EXECUTION]: The generated 'bootstrap.ts' script performs network operations using the Vapi API to create and update voice assistants. This is the intended primary purpose of the skill. It also uses 'bun run' for local task execution.
[CREDENTIALS_SAFE]: The skill handles a 'VAPI_API_KEY' but demonstrates safe management practices by instructing the user to store secrets in a '.env.local' file and providing a '.gitignore' template that explicitly excludes local environment files and secret-bearing directories from version control.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface (Category 8). It ingests untrusted data from a 'ROUGH_DRAFT.md' file and interpolates its content directly into generated system prompts ('body.md') and TypeScript code.
Ingestion points: Content is read from 'ROUGH_DRAFT.md' at the project root.
Boundary markers: None; content is extracted based on Markdown headers and distilled into prompt templates.
Capability inventory: The skill generates a 'bootstrap' script capable of making authenticated network requests to 'api.vapi.ai' and writing numerous files to the local file system.
Sanitization: There is no evidence of sanitization or validation of the input draft content before it is placed into executable code or agent system instructions.

vapi-bootstrap-framework