claude-software-3-0-marketplace
CLAUDE.md2 starsChanged 5 months agoMITView on GitHub
Never shipped; this configures Claude Code for whoever develops the registry. The goal is a tight, high-signal harness for a repo whose only payload is one JSON file. - settings.json — the harness wiring: - includeCoAuthoredBy: false + empty attribution → human-authored commits/PRs (global policy). - permissions.allow → read-only/maintenance Bash (git read verbs, jq, find, grep, cat, the gate runner, claude plugin validate, bun lint/format) so routine work doesn't prompt. - skillOverrides → disable skill families irrelevant to a docs/JSON…
- Reads credentials
In this repository11 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- .claude-plugin/CLAUDE.mdCLAUDE.md
- .claude/teams/CLAUDE.mdCLAUDE.md
- docs/CLAUDE.mdCLAUDE.md
- .github/CLAUDE.mdCLAUDE.md
- site/CLAUDE.mdCLAUDE.md
- add-pluginSkill
- list-pluginsSkill
- remove-pluginSkill
- update-pluginSkill
- vet-pluginSkill
The file
We split this CLAUDE.md into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
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
- Distinctive: line 3
Never shipped; this configures Claude Code for whoever develops the registry.
- Very specific: line 21
All read the hook payload from stdin (JSON) via python3.
≈ “Common” is an estimate: at least 1 in 5 of the 27,641 CLAUDE.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
.claude/ — the author-only harness
Never shipped; this configures Claude Code for whoever develops the registry. The goal is a tight, high-signal harness for a repo whose only payload is one JSON file.
Map
settings.json— the harness wiring:includeCoAuthoredBy: false+ emptyattribution→ human-authored commits/PRs (global policy).permissions.allow→ read-only/maintenance Bash (git read verbs,jq,find,grep,cat, the gate runner,claude plugin validate, bun lint/format) so routine work doesn't prompt.skillOverrides→ disable skill families irrelevant to a docs/JSON registry (frontend, backend, postgres/db, and the instinct-learning set).hooks→ see below.
hooks/marketplace-guard.sh(PreToolUseWrite|Edit|MultiEdit) — blocks badmarketplace.jsonedits (invalid JSON, renamed marketplace, forbiddenversion/sha/commit, secret tokens).guard-commit-author.sh(PreToolUseBash) — blocks agit commitwith Claude/Anthropic author or an AI attribution trailer.json-format.sh(PostToolUseWrite|Edit) — pretty-prints edited valid*.json; no-op otherwise.- All read the hook payload from stdin (JSON) via
python3; exit2blocks,0allows.
skills/— the agent-driven manage-plugins workflow:add-plugin/—/add-plugin <repo>: vet → edit manifest + README + CHANGELOG → gates → PR. Its deterministic core lives inscripts/:{vet-candidate,add-entry,update-entry,remove-entry,sync-readme,lib}.sh(shared by all four skills).vet-plugin/—/vet-plugin <repo>: read-only preflight (go/no-go + blockers).update-plugin/—/update-plugin <name> [--repo …] [--name …]: refresh / repoint / rename an existing entry.remove-plugin/—/remove-plugin <name>: drop an entry. Seedocs/adding-plugins.md.
agents/— the spawnable worker + team agents:plugin-onboarder.md— read-only worker the skills delegate to: fetches + vets a candidate repo and curates its description/keywords; never edits the manifest.registry-dev-*.md(9) — theregistry-devengineering team's roles (auto-discovered); seeteams/below.
teams/— author-only spawnable agent teams (dev-only; never shipped):registry-dev/— the engineering team that implements the brainstorm roadmap (docs/plan/) behind the gate suite.TEAM-BRIEF.md+BACKLOG.md+README.md; entry pointregistry-dev-manager. Its ideation counterpart isdocs/brainstorm/; the handoff isdocs/teams.md.
rules/marketplace-dev.md— path-scoped (.claude-plugin/**) restatement of the manifest contract; loads only when the manifest is touched.agentline.json— statusline (model · branch · PR · context% · tokens · cost) for the agentline statusline.
MCP
.mcp.json (repo root, committed) declares context7 (docs lookup) and github servers. Secrets
are not in the file — they come from env (CONTEXT7_API_KEY, GITHUB_PERSONAL_ACCESS_TOKEN).
Per-machine plugin/MCP enablement lives in the git-ignored .claude/settings.local.json.
Mirror to gates
What the hooks enforce interactively, tests/gates/ enforces in CI — edit them together so local and
CI never disagree.
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.

