Guide · Agent skill
How to use the Agent Doc Discipline agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
Agent Doc Discipline is a skill file for AI coding agents that sets writing rules for documents other agents will read, such as CLAUDE.md files, glossaries, specs, tickets and skill files. It is part of the oh-my-claudecode project and is published under the MIT licence. This guide explains what the skill changes in an agent's writing, how to install it, how to trigger it, and where it does not belong, so you can decide whether it fits your setup.
Agent Doc DisciplineYeachan-Heo/oh-my-claudecode · MIT Download the skill Quick start
- 1Download the ZIP. Get the agent-doc-discipline ZIP from the directory page. It contains SKILL.md.
- 2Unzip into a skills folder. Unzip it into ~/.claude/skills/ for personal use, or into the project's .claude/skills/ folder to scope it to one project.
- 3Or upload in the Claude apps. In the Claude apps, open Settings, go to Capabilities, then Skills, and upload the skill.
- 4Confirm your runtime loads it. Check that your agent runtime is oh-my-claudecode or another SKILL.md-compatible runtime that loads the file.
- 5Ask for a document edit. Describe the agent-facing document you are writing, such as a CLAUDE.md or spec, and say another agent will read it.
- 6Review the result. Check that the document states the why behind each rule, numbers its multi-step procedures, and gives the next action at the end.
What the skill makes the agent do
When an agent writes or edits a document that other agents will consume to act, this skill applies a set of writing rules. The test a document must pass is that a fresh agent session with no chat history can act on it by reading alone.
The rules ask for conditions that can be checked as pass or fail, each stated with its reason. They ask for numbered steps before reference material, one authoritative place for each rule or definition, and no copied lookups such as package scripts or directory listings that would go stale. Each step should end on a completion criterion, so a reader can tell when it is done.
- Rules are checkable and include a why next to them.
- Procedures of three or more steps are numbered.
- Each meaning lives in one place; duplication is treated as future drift.
- Documents do not restate what can be found by looking at the environment.
- Before reporting a document change done, the agent checks that a fresh session could act on it without asking a human.
Which documents the skill covers
The skill names five surfaces: CLAUDE.md, GLOSSARY.md, docs/standards/, design-system/ and .omc/skills/. It also covers specs, tickets and skill files in general.
The skill is mandatory when drydock generates surface seeds, and when the launch C5 sediment pass writes a lesson into a slot. For any other edit to an agent-facing document it is opt-in, though the source says such edits should still survive these rules.
Install the skill
You can install the skill in either of two ways. Download the ZIP from the directory page and unzip it into ~/.claude/skills/ to make it available for your user account, or into a project's .claude/skills/ folder to scope it to that project.
In the Claude apps you can instead upload the skill under Settings → Capabilities → Skills.
- Personal install: unzip into ~/.claude/skills/.
- Project install: unzip into the project's .claude/skills/ folder.
- Claude apps: upload under Settings → Capabilities → Skills.
Check what you need before you start
The skill requires oh-my-claudecode or a SKILL.md-compatible agent runtime that loads this skill file. The package contains a single file, SKILL.md. It runs no scripts and uses no network access.
- oh-my-claudecode, or another agent runtime that reads SKILL.md files.
- The skill package contains SKILL.md only.
Trigger the skill with natural requests
The skill is described to the agent by its name and purpose, so you can ask in plain language. Name the kind of document you are writing or editing and say that another agent will read it. The source does not list fixed trigger phrases, so the examples below describe the kind of request rather than exact wording the skill requires.
Mandatory triggers come from drydock seed generation and the launch C5 sediment pass. Outside those, ask for the skill when you edit an agent-facing document.
- Ask the agent to draft or revise a CLAUDE.md so a new session can follow it.
- Ask it to tighten a spec or ticket that other agents will act on.
- Ask it to write or edit a skill file that other agents will load.
- Ask it to clean up a glossary or shared standards file for agent use.
Know when not to use it
The skill is for documents that agents consume. Documents written only for humans are not held to this discipline. These include docs/business/ narrative, ADR decision stories and READMEs for human onboarding. The source says these may follow the rules where that costs nothing.
The rules are a style constraint, not a safety one. Projects that do not use the oh-my-claudecode surfaces or drydock and launch workflows will not hit the mandatory triggers, and the rules may add process they do not need.
- Business narrative in docs/business/.
- ADR decision stories.
- Onboarding READMEs for humans.
- Projects without oh-my-claudecode, drydock or launch workflows.
Safety notes on scripts and network
The skill runs no scripts and uses no network. Its effect is limited to how the agent writes documents. Because it is a style constraint, do not rely on it to enforce security or access rules; review the documents the agent produces as you would any other change.
What people use it for
- Drafting a CLAUDE.md that a new agent session can follow without chat history.
- Tightening a spec or ticket so each step has a completion criterion.
- Writing a skill file that other agents will load.
- Keeping a glossary or standards file free of duplicated definitions over time.
- Recording a lesson from a launch C5 sediment pass in the right slot.
Frequently asked questions
Is the agent doc discipline skill free?
The skill is published under the MIT licence, according to this directory listing. The source does not mention pricing for the skill itself, so check the licence file in the repository for the full terms.
What do I need to run agent-doc-discipline?
You need oh-my-claudecode or a SKILL.md-compatible agent runtime that loads the skill file. The skill runs no scripts and uses no network.
How do I install agent-doc-discipline?
Download the ZIP from the directory page and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. In the Claude apps you can upload it under Settings, then Capabilities, then Skills.
Does the skill run automatically on every document?
No. It is mandatory when drydock generates surface seeds and during the launch C5 sediment pass. For other agent-facing edits it is opt-in.
Can I use this for human-only documents?
The skill says human-only documents such as business narrative, ADR decision stories and onboarding READMEs are not held to this discipline. They may follow the rules where that costs nothing.
How does it compare with minimal-code-discipline?
The source describes minimal-code-discipline as the companion skill. That one disciplines code, while agent-doc-discipline disciplines documents agents read.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you maintain CLAUDE.md files, specs, shared agent instruction files or skill files in the oh-my-claudecode workflow, or in drydock and launch processes. Skip it for human-only documents or for projects that do not use these surfaces, since its mandatory triggers will not apply there and the rules may add unneeded process. Start by installing it, then make one agent-facing edit and check the result against its verification list.
Download the skill