The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses ollama CLI commands and curl to manage local model runtimes, pull model weights, and interact with the local API on 127.0.0.1:11434. Standard commands like ollama list, ollama ps, and ollama show are used for environment verification.
[EXTERNAL_DOWNLOADS]: Fetches model weights and documentation from official Ollama infrastructure (ollama.com). It also references resources from the vendor's own domain (clawic.com) for documentation and updates.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data including model responses and retrieved chunks for RAG workflows, which could contain malicious instructions.
Ingestion points: Model outputs from the local API (documented in api-patterns.md) and retrieved text chunks from the local corpus (documented in embeddings-and-rag.md).
Boundary markers: The skill recommends using format: "json" and non-streaming responses for deterministic parsing to mitigate accidental instruction obedience.
Capability inventory: Uses curl for network requests and the ollama CLI for model creation and process management.
Sanitization: Recommends validating parsed JSON results before taking downstream actions and using explicit model tags to prevent drifting capabilities.

Ollama