department-brief
Skill0 starsChanged 29 days agoMITView on GitHub
Write or review the brief for a department or role an agent works in: what it is responsible for, what it must not touch, where it works, who approves its output and how it reports, in ten lines or fewer kept in the repo. Use when setting up a department or role, when a session is about to work in an area with no brief, or when asked to review a brief.
In this repository12 other files it gives its agents.
- templates/CLAUDE.mdCLAUDE.md
- accessibility-basicsSkill
- approvalsSkill
- backlog-itemSkill
- component-statesSkill
- forms-feedbackSkill
- grill-before-buildSkill
- handoffSkill
- layout-and-hierarchySkill
- release-checklistSkill
- theme-token-auditSkill
- ui-review-checklistSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 6 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
Use when a team hands a stream of work (product, development, operations, docs, support) to Claude Code sessions and each session needs to …
- Distinctive: line 19
No secrets, tokens, customer data or private contact details in a brief.
- Distinctive: line 45
Read the brief before the first change.
- Distinctive: line 49
No brief yet: draft one from what you can see (folders, owners files, existing tasks), mark it as a draft, and ask the lead to approve it …
- Distinctive: line 54
Check that each part is present and concrete.
- Distinctive: line 69
When writing: the brief as a file in the diff, and the questions you could not answer from the repo, each with your recommended answer.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill 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
Department brief
Use when a team hands a stream of work (product, development, operations, docs, support) to Claude Code sessions and each session needs to know where its job starts and stops. A brief is what you would tell a new hire on day one, and nothing more. Pair it with backlog-item (the work itself) and approvals (the human yes).
Ground rules
- Ten lines or fewer. A long brief is not read; detail belongs in the code, the docs and the tasks.
- Kept in the repo next to the work: a
CLAUDE.mdin that folder, ordocs/roles/<department>.mdthat sessions are pointed to. Not only in a chat or a page that one person can open. - One lead per department owns the brief, named by role ("the docs lead"). A session proposes changes to it; the lead approves them.
- No secrets, tokens, customer data or private contact details in a brief.
The five parts
- Responsibility: what the department delivers and for whom, in one or two lines.
- Must not touch: other departments' areas, production settings, money, anything public. Name them concretely
("the payment code in
src/billing/"), not "be careful". - Where it works: repos, folders, branch prefix, and the commands that check its work.
- Who approves: which role reviews its output, and which of its actions always need a human yes (see approvals).
- How it reports: where the result goes (the pull request, the task, the status file) and what a report holds: what works, what is not done, what could not be checked.
Example
# Docs department
Responsible for: the public docs site (guides and API reference) for developers who integrate the orders service.
Must not touch: application code outside `docs/`, the site's deploy settings, the pricing page.
Works in: `docs/` on `docs/<short-name>` branches; check with `npm run docs:build` and the link checker.
Approves: the docs lead reviews every pull request; publishing the site is a human yes.
Reports: in the pull request: pages changed, what was checked in the preview, open questions.
Asks first: anything that changes what the product promises (limits, prices, guarantees).Using a brief in a session
- Read the brief before the first change. If the task needs something the brief rules out, stop and ask; do not widen the brief yourself to fit the task.
- One session, one task, one department. When the work moves to another department, say so and switch brief rather than drifting across everything in one long session.
- No brief yet: draft one from what you can see (folders, owners files, existing tasks), mark it as a draft, and ask the lead to approve it before relying on it.
Reviewing a brief
Check that each part is present and concrete. Defects worth reporting:
| Defect | What it looks like | Fix |
| No limits | "Must not touch" missing or "be careful with production" | name the folders, settings and actions |
| Nobody approves | "the team reviews", "ask someone" | one role per kind of output |
| Grown into a manual | over ten lines, steps for one particular task | move the steps to the task or the docs |
| Stale | folders or commands that no longer exist | run the commands, check the paths |
| Overlap | two briefs claim the same folder | one owner; the other brief says "ask the X lead" |
| Silent reports | no place or shape for the result | name the place and the three report lines |
A brief that is complete and current is a valid result: say so, do not pad.
How to report
When writing: the brief as a file in the diff, and the questions you could not answer from the repo, each with your recommended answer. When reviewing: one finding per defect with the line, what is wrong and the replacement line.
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.

