from ClaudeSmalltalk15
Connects an LLM (Claude/OpenAI/etc.) to a live Smalltalk (Squeak/Cuis) image over MCP for browsing, evaluating and modifying Smalltalk code.
This skill exposes a set of MCP-connected tools that let an agent inspect and manipulate a running Smalltalk image (Squeak or Cuis). Use it to evaluate expressions, browse classes and categories, read method source, define or modify classes and methods, run SUnit tests, and perform multi-step code audits via the smalltalk_task orchestration tool.
Use this skill when you need the agent to interact with a live Smalltalk environment: code evaluation, reading or editing methods, exploring class hierarchies, running tests, or completing multi-step maintenance tasks (e.g., audit a class, implement a new class with tests). Prefer smalltalk_task for complex, multi-step workflows and individual smalltalk_* tools for quick lookups or single evaluations.
.smalltalk-mcp.json config, and configuring Claude Desktop are provided. The skill documents best practices: always browse before modifying, test after changes, and use side: "class" for class-side methods.Likely compatible with Claude Desktop and other LLM-integrations that support MCP-style tools (Claude, Anthropic models, OpenAI wrappers, Ollama). The SKILL.md includes examples for Anthropics, Ollama, OpenAI and MQTT transports.
Smalltalk MCP connects an LLM to a live Smalltalk (Squeak/Cuis) image via 13 MCP tools for browsing, evaluating, and modifying Smalltalk code. No bundled scripts — it delegates to an external MCP server (smalltalk_agent_mcp.py) from the parent repo. The SKILL.md is well-structured with clear setup instructions, usage guidance, and a tool reference table, though it lacks formal output contracts and troubleshooting for common failures.
Clean skill with no security concerns. Niche audience — Smalltalk developers who want LLM integration. The skill is well-documented but requires substantial setup (Python + Smalltalk VM + image build + JSON config). No scripts were present to test. The tool reference table is helpful but lacks parameter schemas.
