@2sophia/sophia-motor-skill
v0.5.1
Published
Claude Code skill for the sophia-motor Python library — teaches a developer's Claude how to build agents with sophia-motor (the @tool decorator, MotorConfig, RunTask, subagents, MCP integration, multi-turn chat).
Maintainers
Readme
sophia-motor — Claude Code skill
A meticulous skill that teaches your local Claude Code how to build agents with sophia-motor — the @tool decorator, MotorConfig, RunTask, subagents, in-process MCP integration, structured output, and the production patterns the maintainers ship.
Install (recommended) — via Claude Code marketplace
Inside a Claude Code session:
/plugin marketplace add 2sophia/skills
/plugin install sophia-motorClaude Code clones the 2sophia/skills repo, reads .claude-plugin/marketplace.json, and registers the skill. Update later with /plugin update. Zero filesystem ceremony.
Install (alternative) — via npm
If you prefer npm-driven installs:
npm install -g @2sophia/sophia-motor-skill
SKILL_DIR=$(npm root -g)/@2sophia/sophia-motor-skill
mkdir -p ~/.claude/skills && ln -s "$SKILL_DIR" ~/.claude/skills/sophia-motorThe marketplace path is canonical and avoids the symlink dance — prefer it unless you have a specific reason for npm.
What the skill does
When you ask your Claude (in Claude Code) to write Python code that uses sophia-motor, it loads SKILL.md and follows the conventions documented inside — including the golden rule: when uncertain about a field, default, or signature, read the installed source of the sophia-motor package directly rather than guessing.
The skill covers:
- Installation,
.env, first run Motor/MotorConfig/RunTaskAPI + lifecycle + env-var cascade- Built-in tools (Read, Glob, Bash, …) + the strict guardrail
@tooldecorator +ToolContextfor Python functions exposed as in-process MCP toolsSKILL.mdmounting (the agent skills feature)- Attachments (hard-link / symlink quirks)
- Structured output (
output_data) + generated files (output_files) - Streaming + every chunk type
Chat+motor.console()+motor.interrupt()- Subagents (inheritance + explicit-restrict patterns)
- Multi-provider adapters (Anthropic / vLLM / custom)
- Observability — events, logs, audit dump
- Production patterns — singleton motor, concurrency, chat backends
- Reference tables — full field/env-var/event-type lookup
- 10 known gotchas + recovery recipes
Versioning
The skill version tracks the sophia-motor Python package version it was built against. 0.5.1 is built against sophia-motor==0.5.1. When the installed Python version differs, the skill instructs Claude to inspect the installed source for ground truth — fail-safe by design.
Update
- Marketplace:
/plugin updateinside Claude Code - npm:
npm update -g @2sophia/sophia-motor-skill(the symlink picks up the new content automatically)
Uninstall
- Marketplace:
/plugin uninstall sophia-motor - npm:
rm ~/.claude/skills/sophia-motor npm uninstall -g @2sophia/sophia-motor-skill
License
MIT — see LICENSE in the main repo.
