claude-workbench
CLAUDE.md2 starsChanged 5 days agoMITView on GitHub
This repo is the Claude Workbench plugin: a Claude Code / Cowork plugin that builds and runs a personal knowledge vault ("workbench") for anyone, adapted to their role. It's the public, company-neutral version of a vault design proven in daily use. Keep it neutral: no company names, no real people, no tool instances, no internal URLs in anything shipped. - Skills hold behavior; the script holds structure. The script never decides content. Skills never hand-build the folder tree. Keep that…
In this repository10 other files it gives its agents.
- calendar-syncSkill
- captureSkill
- daily-briefingSkill
- eod-recapSkill
- librarianSkill
- migrateSkill
- nightly-dreamSkill
- schedulesSkill
- setupSkill
- weekly-reviewSkill
The file
We split this CLAUDE.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 43
When a skill's behavior changes, update docs/index.html and the README in the same change, so the explainer never promises something the …
≈ “Common” is an estimate: at least 1 in 5 of the 26,980 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 Workbench (plugin source)
This repo is the Claude Workbench plugin: a Claude Code / Cowork plugin that builds and runs a personal knowledge vault ("workbench") for anyone, adapted to their role. It's the public, company-neutral version of a vault design proven in daily use. Keep it neutral: no company names, no real people, no tool instances, no internal URLs in anything shipped.
Layout
.claude-plugin/ plugin.json + marketplace.json (the repo is its own marketplace)
skills/ one folder per skill; the routines' logic lives here
commands/ read-only gatherers: person, project, vault-search
hooks/ SessionStart hook -> scripts/workbench.py status --hook
scripts/ workbench.py: stdlib-only scaffolder (detect/plan/apply/schedules/status/pin)
blueprints/ base.json (folders, files, routines) + roles.json (role packs)
templates/ folders/*.md -> each folder's CLAUDE.md; pages/*.md -> starter pages
tests/ unittest suite for the scaffolder
docs/index.html the visual explainer (self-contained, no external assets); keep it in step with the skillsHow the pieces fit
- Skills hold behavior; the script holds structure. The script never decides content. Skills never hand-build the folder tree. Keep that split.
- Scheduled task prompts stay tiny. They invoke
workbench:<skill>and fall back to the copy the scaffolder writes to<workbench>/.workbench/routines/<skill>.md. So changing a routine means editing itsSKILL.mdand shipping a version; users pick it up via/workbench:migratewithout touching their schedules. - Ownership table is a contract. The root template (
templates/pages/root-CLAUDE.md), each folder template, and each routine skill all state who writes what. If you change what a routine writes, change all three, plusblueprints/base.jsonwrites. - Safe upgrades depend on the manifest.
applyrecords the sha256 of every file it writes. Plugin-managed files (everyCLAUDE.md, READMEs in Tracking,.workbench/*) update only when untouched; content pages (user_owned: true) are written once and never again.
Conventions
- Template placeholders are
{{snake_case}}and must be supplied bybuild_context. The testtest_no_unrendered_placeholdersfails if one slips through. - Prose in templates and skills: plain, direct, contractions fine, no filler. Write for a smart reader who's busy.
- When a skill's behavior changes, update
docs/index.htmland the README in the same change, so the explainer never promises something the plugin doesn't do. - Bump
versionin both.claude-plugin/plugin.jsonandmarketplace.json, and add aCHANGELOG.mdentry. If a release changes content-page shapes, list the migration under the version so/workbench:migratecan apply it.
Test
python3 -m unittest discover -s tests
claude plugin validate .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.

