quick-work
Skill1 starsChanged 3 months agoMITView on GitHub
Lite version of /work for small tasks where the full spec ceremony would be overhead. Use for single-file edits, simple bug fixes, micro-refactors, or anything you'd estimate at <30 minutes. Trades structure for speed while keeping the critical review step.
In this repository17 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- architectureSkill
- code-reviewSkill
- commandsSkill
- conventionsSkill
- debugSkill
- decisionSkill
- explainSkill
- isolateSkill
- issueSkill
- journeysSkill
- onboardSkill
- safe-commitSkill
- safe-pushSkill
- scaffoldSkill
- todoSkill
- workSkill
The file
We split this skill into the rules it gives an agent
10 rules. About 0 are common, 7 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
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 20
If in doubt, use /work. The full spec costs you 3 extra minutes but saves you debugging cycles.
- Distinctive: line 26
Read or ask for the task. Restate it in your own words:
- Distinctive: line 30
Wait for confirmation. This is the only "ceremony" — it catches misunderstandings in 30 seconds.
- Distinctive: line 36
If the change requires reading a file you don't have context on, read it first.
- Distinctive: line 55
If anything is ⚠ and the user might care, ask before considering this done.
- Distinctive: line 73
No 5-question spec phase
- Distinctive: line 74
No formal plan generation
- Distinctive: line 75
No prediction of all failure modes upfront
- Very specific: line 76
No saving spec to /todo for future sessions
- Very specific: line 78
If you find yourself doing 2+ /quick-work invocations on the same logical task, switch to /work.
≈ “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
/quick-work — Lite framework for small tasks
You are Claude. The user has a small task — single-file edit, 1-2 function changes, a CSS tweak, a typo fix, a 3-line refactor — and the full /work spec ceremony is overkill.
/quick-work keeps the senior discipline (think before generate, review before commit) without the 5-question intake. Estimated total session: 5-15 minutes.
When to use this vs /work:
Use /quick-work when... | Use /work when... |
| Change is <30 lines | Change touches multiple files or is non-trivial logic |
| You can articulate the task in 1 sentence | The task needs constraints written down (no new deps, must be backward-compat, etc.) |
| The failure mode is obvious or trivial | Failure modes need explicit thought (data integrity, race conditions, etc.) |
| You're not committing afterward (just exploring) | You'll commit and need the spec to anchor the commit message |
If in doubt, use /work. The full spec costs you 3 extra minutes but saves you debugging cycles. /quick-work is for when 3 minutes IS the entire task.
Argument (optional): $ARGUMENTS may be the task description. If empty, ask.
Phase 1 — Restate
Read or ask for the task. Restate it in your own words:
"Got it — you want me to [restated task]. Constraints I'm assuming: [stack from project files], no new deps unless asked, change scope limited to [obvious area]. Confirm or correct?"
Wait for confirmation. This is the only "ceremony" — it catches misunderstandings in 30 seconds.
Phase 2 — Generate
Make the change. Show the diff inline (not just the new code).
If the change requires reading a file you don't have context on, read it first. Don't guess at structure.
Phase 3 — Self-check (the discipline that survives)
Out loud, run these three quick checks:
- Does it solve the actual task? (Read the diff against the restated task)
- Does it break anything obvious? (Type errors, missing imports, undefined refs, mismatched function signatures)
- Edge cases worth noting? (Empty input, null, etc. — call them out even if you don't handle them)
Output format:
SELF-CHECK
✓ Solves: <task restated>
✓ No obvious breakage: <X type checks fine, imports resolved, etc.>
⚠ Edge cases noted: <list any you're aware of but didn't handle, with reasoning>If anything is ⚠ and the user might care, ask before considering this done.
Phase 4 — Hand off
End with one of:
"Done. Run
/safe-commitif you want to commit this with the review gate, or justgit add + commitdirectly if you're confident."
OR (if user mentioned a test):
"Done. Run the test to verify before committing:
<command>."
OR (if exploration, not committing):
"Done. Diff is above — let me know if you want to iterate or commit."
What /quick-work does NOT do
- No 5-question spec phase
- No formal plan generation
- No prediction of all failure modes upfront
- No saving spec to
/todofor future sessions
If you find yourself doing 2+ /quick-work invocations on the same logical task, switch to /work. The compounding context cost exceeds the spec ceremony cost.
What it DOES do
- Forces the restatement (the cheapest misunderstanding-catcher there is)
- Forces the self-check (catches the obvious breakage before it reaches a commit)
- Hands off cleanly to
/safe-commitif needed
That's the senior discipline distilled to its irreducible minimum.
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.

