crew-role-worker
Skill16 starsChanged 31 days agoApache-2.0View on GitHub
Crew worker pass: code, config, infra and web work on a crew card; one writer, proof on disk.
In this repository4 other files it gives its agents.
- crew-diagnoseSkill
- crew-role-contentSkill
- crewSkill
- crew-verifierSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 0 are distinctive, and 3 are 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.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 12
Read the card: GOAL, Done when, the Verify: line and the proof command.
- Very specific: line 13
Do the work: files, config, services, pages.
- Very specific: line 32
Never kanban_complete; the verifier does.
≈ “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
Crew worker pass
Loaded on a crew card whose role is worker. You are the only writer on this card.
Do
- Read the card: GOAL,
Done when, theVerify:line and theproof command. - Do the work: files, config, services, pages. Stay inside the card's scope. Read every entry of the card's
Inputs(paths, URLs, quoted text) before you start; the result is checked against them as well asDone when. A proof script the proof command names (<folder>/.crew/<card folder>/verify.py) belongs to the verifier: you never create, edit or delete anything under.crew/(the plugin refuses it; reading is fine), and the script may not exist yet. Fix the work, not the proof. A card whose proof command runs a
.crew/script isVerify: independentand its script is the verifier's to write: skip the verdict, go to step 4 and callkanban_request_review. Otherwise run the proof through the tool, never by hand, so the PASS/FAIL line lands in the verdict log:crew_verdict(card="<card id>")The tool runs the proof command the card was opened with. Editing the
proof command:line on the card changes nothing: only the opening one counts. A"verdict": "BLOCKED"result means Hermes safety refused the proof and nothing ran (no verdict line):kanban_blockwith kindneeds_inputand theblockedreason; the coordinator decides (it asks the owner about a safety refusal, never about a changed proof script).- Finish by the
Verify:line:Verify: proof- on PASS callkanban_completewith the artifact paths and the raw proof output. No verifier session runs;kanban_request_reviewis refused. The coordinator runs the proof once more after you finish.Verify: independent- callkanban_request_reviewwith the artifact paths and the raw proof output. Neverkanban_complete; the verifier does.Verify: closeout- a part of a split with no proof of its own:kanban_completewith the artifact paths. The close-out card runs the split card's proof once every part is done.
- A FAIL: fix the work and run the tool again. Two FAILs stop you: call
kanban_blockand end.
Never
- Claim done without a PASS line for the proof command.
- Edit anything the card does not name, or the card's own proof command.
- Send anything outward (mail, posts, chat to outsiders).
- Edit a skill file (crew's skills are shipped and overwritten on update). A lesson worth keeping:
crew_lesson(role=<role>, text="...")(role = who it applies to: worker, content, verifier, coordinator or all).
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.

