Instruction Tools

The 20 instruction tools are the public surface of the server. Each tool accepts a { request: string; context?: string; options?: object } payload and internally orchestrates skills.

Model class badges describe task complexity, not cost to users — the class tells the orchestrator what tier of model can handle this task well:

Badge	Meaning	Recommended models
Zero-Cost	Solved by zero-cost models — drafts, triage, templates	GPT-4.1, GPT-5 mini
Efficient	Low-cost models — routing, aggregation, fast classification	Claude Haiku 4.5, GPT-5.4 mini
Advanced	Full-capability models — synthesis, security, judgment	Claude Sonnet 4.6, Claude Opus 4.6, GPT-5.4, GPT-5.3-Codex
Cross-Model	Independent cross-model review lane	Gemini 3.1 Pro, Claude Opus 4.6

→ GitHub model comparison · Artificial Analysis

Tool Registry

Start here at the beginning of a new session. These tools orient the model to the project and route subsequent requests.

task-bootstrap Bootstrap: First Contact — Zero-Cost. Onboards the model to a new session and project context.

meta-routing Meta-Routing: Task Router — Efficient. Routes any ad-hoc request to the right specialized tool.

project-onboard Onboard: Project Familiarization — Zero-Cost. Deep-reads codebase structure, conventions, and history.

Input Schema

All tools share the same input schema:

{
  "request": "string (required)",
  "context": "string (optional)",
  "options": "object (optional, tool-specific)"
}