arach/dewey

Overview

This skill generates AI-agent-ready documentation using Dewey. It scaffolds and produces AGENTS.md, llms.txt, docs.json, and install.md so coding agents can consume project context reliably. Use it to make your docs structured, machine-friendly, and resumable for automated workflows.

How this skill works

The skill runs Dewey CLI commands or uses its configuration to inspect your repository and docs folder, then emits agent-focused artifacts. It extracts critical context, entry points, rules, and compact agent summaries, and writes both human and agent-optimized pages (e.g., docs/agent/*.agent.md). It can also audit completeness and score agent-readiness so you can iterate toward a target score.

When to use it

• When you want AGENTS.md or llms.txt generated for an agent workflow
• When you need to make docs machine-readable and self-contained
• When onboarding LLM-based assistants or CI agents to a repo
• When you want an install.md with verifiable setup steps
• When checking documentation completeness or agent-readiness

Best practices

• Keep agent pages dense: facts, tables, explicit values, no prose
• Mirror human docs with agent/ *.agent.md versions next to each page
• Define clear criticalContext and entryPoints in dewey.config.ts
• Include reproducible verification commands for install.md
• Run dewey audit and aim for an agent score above 80 before shipping

Example use cases

• Generate AGENTS.md and llms.txt for an open-source library to hand to an AI code assistant
• Create install.md that CI agents can execute and verify during onboarding
• Add agent-optimized API tables (params, types, errors) for programmatic consumption
• Audit a mono-repo to identify missing docs and enforce conventions
• Produce docs.json for tooling that indexes documentation and enables search