Route review tasks through run_helper with local-model MCP presets.
npx skills add https://github.com/bja2142/vibes --skill local-model-mcp
Please help me install this Agent Skill. Skill: local-model-mcp Source: https://github.com/bja2142/vibes/tree/main/model-mcp/skills/local-model-mcp Command: npx skills add https://github.com/bja2142/vibes --skill local-model-mcp
The Local Model MCP provides a centralized way to obtain second opinions, reuse reusable presets, and perform cross-model comparisons with a local MCP server. It also supports session management, export, and handoffs across models to improve review quality and continuity.
Invoke the local-model-mcp via run_helper with a preset or raw provider to start a review.
High-intent search queries and answers about installing and using this skill.
To get a second opinion, route your task through the run_helper command with a raw provider or preset. This local-model workflow applies cross-model comparisons and preset-driven reviews to generate alternative feedback.
Reusable presets are predefined review configurations like security review or adversarial critique. You apply them through the run_helper command to consistently execute specialized review tasks across different local models.
You manage and hand off sessions by using session management commands within run_helper to inspect, label, export, or move stored sessions. This allows moving a stored session from one provider to another for cross-model synthesis.
Yes, you can perform cross-model comparison by routing tasks through run_helper. The workflow supports calling different providers and synthesizing their outputs to compare responses directly within your local-model environment.
Yes, a local MCP server is required. The skill coordinates preset-driven reviews and session handoffs by routing tasks through the local-model MCP, relying on the server to manage provider calls and stored sessions.
v36372Configure global or project MCP servers in mcp.json for stdio or HTTP connections.
TylerGarlickGenerate structured session summaries and export epistemic outputs as Markdown, JSON, or HTML.
majiayu000Estimate tokens, prune messages, and auto-compact LLM context histories.
fcouryReview code, plans, and approaches using the Codex model.
