shipyard
AGENTS.md66 starsChanged 5 months agoMITView on GitHub
Agent-driven plugin (v4.8.0) for structured project execution: idea → brainstorm → plan → build → ship. Shipyard is authored as a Claude Code plugin (canonical source) and also ships a generated Codex plugin tree so Codex users can install it natively. The Claude Code artifacts are the single source of truth; the Codex tree is a build output (see "Codex support" below). Codex artifacts are generated from the canonical Claude Code artifacts — never hand-edited. - scripts/build-codex.sh transforms .claude-plugin/plugin.json +…
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,774 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository20 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- code-simplificationSkill
- documentationSkill
- git-workflowSkill
- import-spec-fileSkill
- import-specSkill
- infrastructure-validationSkill
- lessons-learnedSkill
- parallel-dispatchSkill
- security-auditSkill
- shipyard-brainstormingSkill
- shipyard-debuggingSkill
- shipyard-executing-plansSkill
- shipyard-handoffSkill
- shipyard-tddSkill
- shipyard-testingSkill
- shipyard-verificationSkill
- shipyard-writing-plansSkill
- shipyard-writing-skillsSkill
- using-shipyardSkill
The file
We split this AGENTS.md into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 69
Codex tree: never edit plugins/ or .agents/ by hand — run scripts/build-codex.sh
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Shipyard
Agent-driven plugin (v4.8.0) for structured project execution: idea → brainstorm → plan → build → ship.
Shipyard is authored as a Claude Code plugin (canonical source) and also ships a generated Codex plugin tree so Codex users can install it natively. The Claude Code artifacts are the single source of truth; the Codex tree is a build output (see "Codex support" below).
Commands
npm test # Full BATS test suite
npm run test:fast # Unit tests only (bats --filter-tags unit)
npm run test:ci # Parallel execution (4 jobs)
bash scripts/check-versions.sh # Verify version sync across files
bash scripts/build-codex.sh # Regenerate the Codex plugin tree from canonical artifacts
bash scripts/check-codex-sync.sh # Verify the committed Codex tree matches the generator
shellcheck --severity=warning scripts/*.sh hooks/*.sh test/run.shArchitecture
commands/ Slash commands (markdown + YAML frontmatter) — Claude Code
skills/ Auto-activating skills (skills/<name>/SKILL.md) — shared with Codex
agents/ Specialized subagents (markdown + YAML frontmatter) — Claude Code
scripts/ State management, version checks, Codex generator, cleanup utilities
hooks/ Plugin lifecycle hooks (SessionStart, TeammateIdle, TaskCompleted, Stop) — Claude Code
test/ BATS test suite + test_helper.bash
docs/ Protocols, guides, state schema, context engineering
.claude-plugin/ Claude Code plugin metadata (plugin.json, marketplace.json) — canonical
plugins/ Generated Codex plugin tree (plugins/shipyard/.codex-plugin/, skills/)
.agents/ Generated Codex marketplace manifest (.agents/plugins/marketplace.json)
.husky/ Pre-commit: check-versions.sh && check-codex-sync.sh && npm testCodex support
Codex artifacts are generated from the canonical Claude Code artifacts — never hand-edited.
scripts/build-codex.shtransforms.claude-plugin/plugin.json+skills/into the Codex tree underplugins/shipyard/and.agents/plugins/marketplace.json.scripts/check-codex-sync.shregenerates into a temp dir and diffs against the committed tree; any drift fails CI. This is what keeps the Codex tree from drifting by hand.- Codex install:
codex plugin marketplace add lgbarn/shipyard, then enable the plugin. - Codex support is best-effort, not parity: skills run natively, but parallel subagent dispatch and lifecycle hooks degrade (see the PRD and the "Using Shipyard with Codex" guide).
Version Sync
Six version strings must match. scripts/check-versions.sh enforces this (also runs in the
pre-commit hook and CI).
package.jsonpackage-lock.json(root +packages[""]).claude-plugin/plugin.json.claude-plugin/marketplace.json(top entry)plugins/shipyard/.codex-plugin/plugin.json(generated — regenerate after a version bump)CHANGELOG.md(top entry)
Conventions
- Files/dirs: kebab-case
- Commits: conventional format —
feat:,fix:,chore:,build:,shipyard(phase-N): - Bash scripts:
set -euo pipefail, ShellCheck compliant (--severity=warning) - Skills: auto-discovered from
skills/*/SKILL.md— no manual registration - Agents: model values are
opus,sonnet,haiku, orinherit - Codex tree: never edit
plugins/or.agents/by hand — runscripts/build-codex.sh - Adding components: see CONTRIBUTING.md
Gotchas
- Pre-commit hook runs
check-versions.sh && check-codex-sync.sh && npm test— all must pass .shipyard/is gitignored — local project state only, never committed- jq >= 1.6 required —
state-read.shexits code 3 if missing; the Codex generator also needs jq - Symlinked
.shipyard/is rejected (prevents writes outside project) - Team locking uses directory-based locks at
${TMPDIR}/shipyard-state-${hash}.lock - CI matrix tests on macOS, Ubuntu, and Windows (WSL); ShellCheck + version + Codex-sync checks run on Ubuntu only
- PR version bump enforced — CI rejects PRs where package.json version <= main branch version
- After a version bump regenerate the Codex tree (
bash scripts/build-codex.sh) orcheck-codex-sync.shwill fail
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

