Borrowing it
Nothing to install: this file belongs to edochi/mdvs. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/edochi/mdvs/main/.claude/skills/ship/SKILL.mdgit clone --depth 1 https://github.com/edochi/mdvsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/edochi/mdvs/ship)<a href="https://agentmods.dev/skills/edochi/mdvs/ship"><img src="https://agentmods.dev/badge/skills/edochi/mdvs/ship/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/edochi/mdvs/ship"><img src="https://agentmods.dev/badge/skills/edochi/mdvs/ship.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00099 | $0.03677 |
| Opus 5.5 | $0.00040 | $0.01471 |
| Sonnet 5.5 | $0.00020 | $0.00735 |
| Haiku 4.5 | $0.00010 | $0.00368 |
Grade A, and why
ship scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 10d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 239 lines — stays where its author put it; the contents beside it link to each section on GitHub.
ship
Why this exists
Agents write code faster than a human can absorb it, so the bottleneck is no longer writing or even verifying (agents increasingly self-verify) but ownership: you cannot steer the next iteration of a system you do not hold in your head, and you cannot feel responsible for decisions you never saw made.
Left unchecked, the loop collapses to "agent writes → coordinator validates →
human rubber-stamps a commit they never read." ship prevents that. It routes
every change through explicit gates the human owns.
ship is the orchestrator. It does not itself explain diffs, run reviews, or
implement code. It sequences those, and — when the human asks for it — invokes
the internalize skill at the ownership gate.
The pipeline
Six stages. The gate owner is who must act before the stage advances.
| Stage | What happens | Gate owner |
|---|---|---|
| 1. Design | Surface every real decision as a decision, with a recommendation. Always run the ship-designer agent before presenting the forks to the human — it grounds the options and hunts for the flaw that sinks the design. Then the coordinator double-checks its every load-bearing claim against the real code (agent output is a hypothesis) before relaying or acting. The agents NEVER decide a fork. |
Human decides each fork. |
| 2. Plan | Draft the implementation agent's prompt with the decisions baked in, STOP triggers enumerated, the anti-patterns forbidden, and the downstream updates the change needs (spec, book, example_kb — see the code-editing skill). Show the scope to the human for anything non-trivial. |
Coordinator drafts; human sees scope. |
| 3. Implement | One ship-implementor agent writes the change and, before returning, runs the full local gate itself (tests + clippy + fmt + ast-grep) and reports the results. Never parallel implementors. (Parallel read-only investigation is fine and encouraged.) |
Coordinator runs it. |
| 4. Validate & review | Coordinator verifies the output against the real code (agent output is a hypothesis — check every non-trivial file:line / API claim) and re-runs the gate, classifying any failure as pre-existing (prove it) vs real. Then ship-reviewer agent(s) attack the change, one lens each; scale their number and lenses to scope: 1 for a small change; 2–3 distinct lenses (correctness, layering/API, test-quality/bug-immunity) for complex or coupled work, and the simplicity/idiom lens on every new module or any diff over a few hundred lines: is this the least code that does the job, what would a senior Rust reviewer delete, is serde / the type system / std doing the work instead of hand-written code. Implementor + reviewer(s) form a persistent wave team that loops fix→re-review until the reviewers are clean (see below). |
Coordinator routes + verifies. |
| 5. Internalize | Opt-in — skipped unless the human explicitly asks for it. When requested: commit locally first (human-approved) so the explainer anchors to a stable SHA, then dispatch ship-internalizer on that commit's diff. The human reads the explainer and takes the judgment quiz. This is where approver becomes participant. |
Human internalizes (when requested). |
| 6. Save | Commit, push, open or update the PR, or move to the next wave — each on the human's explicit word. | Human authorizes commit / push / PR. |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 10d ago First seen · 239 lines · 99 tokens per session scan A 2dd6a595a77f
ship is a skill published in the GitHub repository edochi/mdvs (27 stars, last pushed 11d ago), licensed MIT. It adds 99 tokens to every session and 3,677 once invoked, about $0.0004 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-10-05.
Other skills, from other repositories
esm
Comprehensive toolkit for protein language models including ESM3 (generative multimodal protein design across sequence, structure, and function) and ESM C (efficient protein embeddings and representations). Use this skill when working with protein sequences, structures, or function prediction; designing novel…
llmwiki
Build and use a persistent, citation-traceable knowledge wiki from documents, research papers, notes, or project documentation. Use for llmwiki projects, retrieving reusable evidence for an agent task, or compiling a corpus for repeated questions. A one-off file read or live web search does not need a wiki.
crit-story
Author a crit story and continue the interactive review loop only when the user explicitly invokes crit-story or directly asks you to generate a crit story. Do not infer this skill from generic review, PR, or diff-review requests.
processing-markdown
Processes Markdown files using mq, a jq-like query language for Markdown. Use when the user mentions Markdown processing, content extraction, document transformation, or mq queries.
web-scraping
Fetches web pages and extracts structured data using mq's toolchain (mq-crawl for fetching/crawling/JS rendering, mq for HTML-to-Markdown selector-based extraction, http() for in-query requests). Use when the user wants to scrape a URL, pull structured data out of a webpage, crawl a site, or turn HTML into…
embeddings-search
Use when choosing an embedding model, chunk size, or query form, when semantic search returns irrelevant results, when adding hybrid BM25+vector or a reranker, or when a retrieval change needs a number (recall@k, nDCG, MRR). NOT operating the store — index tuning, quantization (that is vector-db) — nor the…