The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches model lists from official provider endpoints (OpenAI, Anthropic, xAI, Google, Groq) to verify model existence.
[COMMAND_EXECUTION]: Uses local tools like git, rg, uv, and make to analyze the codebase, identify required changes, and run tests.
[INDIRECT_PROMPT_INJECTION]: The skill ingests a user-provided model id which is then interpolated into shell commands and Python scripts for capability probing and SDK checks.
Ingestion points: User input for provider and model id at the start of the workflow.
Boundary markers: The instructions do not explicitly define delimiters for the model id when used in subsequent shell or Python execution contexts.
Capability inventory: The skill utilizes Bash, Write, Edit, and WebFetch capabilities.
Sanitization: A validation step is included (Step 1) that verifies the user-provided ID against official provider listings before proceeding, acting as a safeguard.
[DYNAMIC_EXECUTION]: Instructions guide the agent to generate and run a local Python script (local-notes/probe_<model>.py) to compare model capabilities.

add-new-model