Guide · Agent skill
How to use the Drydock shared repo harness agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
Drydock is an agent skill: a folder of instructions that Claude loads so it can carry out a specific job. This one sets up a shared working environment in a code repository, so people and AI agents read the same written conventions and vocabulary instead of relying on chat history. This guide explains what the skill does, how to install it, how to ask for it, and what to check before and after a run.
Drydock Shared Repo HarnessYeachan-Heo/oh-my-claudecode · MIT Download the skill Quick start
- 1Download the ZIP. Get the drydock skill ZIP from the page you are reading.
- 2Unzip it into the right folder. Place the skill in ~/.claude/skills/ for all projects, or in the repo's .claude/skills/ folder for one project.
- 3Open the repo. Start your agent session inside the repository you want to set up, so the skill can inventory the existing files.
- 4Ask for a run. Request the drydock skill in plain language. Use the --check option if the repo already has this harness.
- 5Review the report. Read what was created, deferred or skipped, and confirm the document language before you commit the changes.
What the drydock skill makes the agent do
When run in a code repository, drydock first inventories what already exists, such as CLAUDE.md, GLOSSARY.md, docs folders and design-system folders. It then seeds two starter files, CLAUDE.md and GLOSSARY.md, so that every contributor and agent starts from the same design language.
Other parts of the shared structure are not created up front. Each one is created the first time it is needed. The skill also detects the document language, migrates recognizable legacy glossary files, and reports what was created, deferred or skipped.
- Context: GLOSSARY.md, docs/business/, docs/adr/ and the OMC wiki
- Rules: CLAUDE.md as a thin entry point, plus docs/standards/
- Tools: .omc/skills/, .mcp.json and scripts/
- Standards: design-system/ (tokens, components, patterns) and docs/standards/
Install the drydock skill
Download the ZIP for the skill from the page you are reading. Unzip it into one of two places. Use ~/.claude/skills/ to make the skill available in all your projects. Use a project's .claude/skills/ folder to make it available only in that repo and to anyone who clones it.
If you use the Claude apps instead of Claude Code, upload the skill under Settings, then Capabilities, then Skills.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the repo's .claude/skills/ folder
- Claude apps: Settings → Capabilities → Skills, then upload the skill
Trigger the skill with a natural request
Open the repo you want to set up and ask the agent to run the drydock skill. The skill's argument hint is [--check], so you can ask for an audit instead of a full setup. A full run is meant to happen once per repo. After that, use the check mode.
Be specific about the repo you are in. The skill reads the files that are present, so running it from the right folder matters.
- Full setup: ask the agent to run drydock on this repo
- Audit an existing harness: ask the agent to run drydock with --check
/drydock --checkUnderstand what a run reports
The skill reports its map of existing files before it changes anything. It then resolves the document language. It first checks the documentLanguage key in the YAML frontmatter at the top of GLOSSARY.md. If that is missing or invalid, it looks at CLAUDE.md and then README.md. If the answer is still unclear, it asks you one batched question instead of guessing.
If a legacy glossary CONTEXT.md exists and there is no GLOSSARY.md, a full run renames it to GLOSSARY.md, using git mv when the file is tracked. An unrelated CONTEXT.md is left alone.
- Existing AGENTS.md and CLAUDE.md are extended in place; the skill never creates both fresh
- Chinese documents must resolve to zh-Hans or zh-Hant; bare zh is asked about
- --check audits the legacy file in place and does not rename or rewrite anything
Safety notes before you run it
The source states that this skill does not run scripts and does not use the network. It requires nothing else to be listed.
A full run can rename files and rewrite legacy CLAUDE.md references, such as changing a pointer to CONTEXT.md so it points to GLOSSARY.md. Do not use a full run on repos where you do not want existing CLAUDE.md or CONTEXT.md files renamed or rewritten. In that case, use --check only, or commit your work first so you can review the changes with version control.
- Runs scripts: no
- Uses network: no
- Full runs can rename and rewrite legacy CLAUDE.md and CONTEXT.md content
What people use it for
- Setting up a new shared repo so the agent and teammates read the same conventions from CLAUDE.md and GLOSSARY.md
- Moving project conventions out of chat history and people's heads into files
- Auditing an existing harness for drift from the expected layout using --check
- Migrating an older glossary CONTEXT.md to GLOSSARY.md during a full run
Frequently asked questions
How do I install the drydock skill?
Download the skill ZIP from the page and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder for one repo. In the Claude apps, upload it under Settings, then Capabilities, then Skills.
Is the drydock skill free?
The skill is published under the MIT licence, as listed on its page. The source does not describe any pricing, so check the page for the current licence terms.
What does drydock create in my repo?
It seeds two files on the first run: CLAUDE.md and GLOSSARY.md. Other parts of the structure, such as docs/adr/ or design-system/, are created only when they are first needed.
Does drydock need any tools or network access?
The source lists no requirements, says the skill does not run scripts, and says it does not use the network.
What is the difference between a full run and --check?
A full run inventories and seeds the harness, and can migrate legacy glossary files. --check audits the repo for drift from the layout and does not rename or rewrite anything.
Should I use drydock for a small throwaway prototype?
No. The source says it is not suited to throwaway prototypes with no collaborators, because the setup adds files with little benefit.
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
Drydock suits teams starting a repo where humans and AI agents will both work from the same written context, and maintainers who want a repeatable audit with --check. Skip it for throwaway prototypes, for repos already set up with this harness (use --check there), and for repos where a full run could rename or rewrite your existing CLAUDE.md or CONTEXT.md files.
Download the skill