The Agent Skills Directory

[DATA_EXFILTRATION]: The skill manages authentication by accessing the local ~/.build-host/credentials.json file.
This access is required for interacting with the build.host API (e.g., deploying projects, updating environment variables) and is consistent with the vendor's own infrastructure.
The skill includes explicit safety rules to prevent printing or leaking any API keys or secrets.
[PROMPT_INJECTION]: The skill operates on user-provided project directories and git repositories, creating an indirect prompt injection surface.
Ingestion points: Local project files and git repository contents during the stack inspection and implementation steps.
Boundary markers: None explicitly defined in the instructions to isolate processed data from the agent's instructions.
Capability inventory: The skill can edit local files, perform network API requests to build.host, and execute HTTP verification checks.
Sanitization: The skill emphasizes operational safety and requires confirmation for destructive actions, though it does not specify content sanitization filters for the project files it reads.

realtime