Coding-agent integration

Your coding agent already lives in the repo. This guide wires it to contentbit so "write a blog post about X" produces a validated document, and "audit my content" reports real findings from document stats.

The integration is deliberately thin: skill files contain no block schemas. They teach the agent to fetch live data from the CLI — the authoring guide from instructions, health data from stats, diagnostics from validate — so when you add a custom block, every agent picks it up on the next run. Nothing to regenerate, nothing to drift.

A session looks like this. The failed validation in the middle is the loop doing its job:

content/dark-mode.md:18:1 error CB_ROW_COLUMNS:::comparison rows require 3 columns (label | left | right). Found 2.hint: Format: - label | left | right

1 file(s): 0 errors, 0 warnings

Install

contentbit init installs the integration automatically (skip with --no-agents). In an existing project, or after upgrading the CLI:

contentbit agents

This manages two things:

• AGENTS.md — a fenced block between <!-- contentbit:start --> and <!-- contentbit:end --> markers with compact instructions any agent reads (Cursor, Codex, Copilot, ...). Your content outside the markers is never touched; re-running refreshes only the block.
• Claude Code skills — when a .claude/ directory exists (or forced with --claude when it doesn't), two skills land in .claude/skills/:
  • contentbit-author — the writing loop: fetch the authoring guide, write, validate, fix diagnostics until exit 0.
  • contentbit-audit — the health check: run stats across content, flag validation issues, thin documents, and block-less pages, in that order.

Pass --no-agents-md to manage only the skills.

What the agent does

The author skill follows the same loop you would run by hand:

# 1. the live guide, including your custom blocks
contentbit instructions --audience llm --registry ./blocks/registry.ts

# 2. write the document, then validate until clean
contentbit validate content/new-post.md --registry ./blocks/registry.ts

Both skills read the content:check script in your package.json to discover the project's content glob and --registry flag, so they adapt to your layout without configuration.

The "never guess props" rule in the guide is satisfiable: generated authoring instructions document every prop from the block's schema — name, type with enum values spelled out, required or optional, default, and description — so optional props an example doesn't use are still discoverable. See LLM authoring.

The audit skill leans on stats, which prints JSON and always exits 0:

contentbit stats "content/**/*.md" --registry ./blocks/registry.ts

Each entry carries a heading outline with per-section word counts, block usage by name, link domains, and a validation summary — enough to rank what needs attention without opening a single file. Full field reference and recipes: Document stats & auditing.

Plug contentbit into your own skills

You likely already have skills for your content workflows — a blog-post skill, a changelog skill. Add contentbit as a context source and a quality gate with three lines:

Before writing, run `contentbit instructions --audience llm --registry <path>`
and follow it exactly — never guess block syntax. After writing, run
`contentbit validate <file> --registry <path>` and fix every diagnostic
(`file:line:col severity CODE message`, with `hint:` lines) until it exits 0.

For data-informed writing, feed contentbit stats output in as well — for example, have the skill check sibling documents' outlines and block usage so new content matches the structure of what already works.

This works in any agent that can run shell commands. The CLI is the interface; skills, rules files, and prompts are thin adapters over it.