OntoShip plugin
A gitmark plugin for Claude Code: a project knowledge base from markdown and git with FTS5 search, a link graph and dev-flow
- 126 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The /ship command moves a change through dev and main branches to production and runs agents with write access.
- The indexer reads every markdown file in the repository and writes a local index.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/ontoshipDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The /ship command moves a change through dev and main branches to production and runs agents with write access.
- The indexer reads every markdown file in the repository and writes a local index.
This is third-party code. Review the repository files before installing.
What it does
OntoShip turns project documentation into a knowledge base of plain markdown files, README indexes and git. The gitmark plugin carries three skills and five commands. The `kb-search` skill searches the base instead of blind grep, `kb-curate` sets light ontology rules for documents and `dev-flow` runs the development loop: research, tasks, spec, an isolated git worktree, implementation, tests, independent review and shipping. Commands: `/kb` searches and answers from the top hits, `/kb-map` builds an HTML link graph, `/doc` writes or updates one document, `/onto-doc` builds the whole base with parallel agents and `/ship` runs dev-flow. The core is `gitmark`, a single Python standard library file with no dependencies: SQLite FTS5, trigram and fuzzy search, frontmatter queries and a linter.
Who it is for. Teams keeping project documentation in markdown who want fast agent search over it and order in the docs without a separate service.
Good fit when
- The agent digs through docs with grep and does not find the right document on the first try.
- A repository has no documentation and it must be built from scratch with parallel agents.
- A map of document links in one HTML file is needed for a team discussion.
- A feature must reach production through a loop with a spec, review and tests.
Not a fit when
- Knowledge lives in a wiki or Notion: the plugin is built for markdown in the repository.
- You need semantic embedding search: gitmark goes without embeddings.
- The team is not ready for an automatic loop to the main branch through /ship.
Example request
/kb how is the billing webhook signature verifiedLimitations
gitmark needs Python 3.7 or newer with SQLite FTS5, and the trigram tokenizer requires SQLite 3.34 or later, otherwise fuzzy and substring search are unavailable. The index lives in `.gitmark/`, which is added to `.gitignore`. The `/ship` command moves a change through a merge request into the dev branch and then main, so permissions and the branch process must be set up beforehand. The author notes that the `destructive-guard` safety hook moved to a separate repository. The author's benchmark of 198 questions gave 69 percent top hits against 30 for grep.
How to disable. Run `/plugin uninstall gitmark@ontoship` or delete the copied `skills` and `commands` from `.claude/`. The `.gitmark/` index directory is deleted by hand.
FAQ
How do I install the plugin?
With `/plugin marketplace add vakovalskii/ontoship` and `/plugin install gitmark@ontoship`. Without the marketplace, copy `skills` and `commands` into the repository's `.claude/`.
Do I need embeddings or a database?
No. Markdown is the source of truth and the SQLite FTS5 index is rebuilt from files with `gitmark index`. There is no service, cloud or embeddings.
How should I write a /ship request?
Describe what to do, why and the done criteria, and name the service or file. A phrase like `/ship fix search` is weak because the agent has to guess the goal.
What does /onto-doc do?
Builds or rebuilds all documentation: splits the repository into areas, runs one curator per area in parallel, then lints, indexes and builds the graph.
Does search work with Cyrillic?
Yes. The trigram tokenizer and fuzzy search find typos and substrings, including Russian text.
Discussion
No questions or reviews yet.
Related
PluginsA Claude Code plugin for phased project execution: planning, parallel agents with TDD, security and quality gates, then shipping.
A Claude Code plugin marketplace by Frad LEE: Git and GitFlow, GitHub, refactoring, memory, research, image generation and interface work
Tasks, docs, specs and decisions live as markdown files in your repository, and the agent reads them through the CLI and MCP
Claude Code plugins: plan, work and review commands, subagents, hooks and writing skills in a practical minimalism spirit.
Plugins for git, Python and accessibility in one place: 400+ skills, subagents and hooks against dangerous agent commands.
A Claude Code plugin that generates and maintains CLAUDE.md: a 0 to 100 quality score, a 150-line cap and stale reference checks
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| OntoShip (this card) | no | 126 | ||
| RepoWiki | no | 306 | generates a wiki from a repository; OntoShip searches your markdown documentation and drives the dev-flow loop | |
| LLM Wiki Memory | no | not required | 130 | agent memory shaped as a wiki; OntoShip relies on README indexes and git with no service |
README in short
The README shows a knowledge base at four zoom levels, from the document graph to a single frontmatter line. The optional ontology covers document types, properties and typed links.