task-classification
Skill0 starsChanged 5 months agoMITView on GitHub
Use at the very start of a task, before anything else, to decide how much process it deserves — chat / trivial edits / simple docs get done immediately; design / build / exploration enter the loop. Judge by blast radius, not by how long the request was. Part of meta-workflow (Step 0).
1 other project uses a version of this.
Adopted by
Most recently taken up by
- josix/agent-flowtheir own version, first seen Oct 20, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- josix/agent-flow★ 8same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository6 other files it gives its agents.
- clarify-intentSkill
- execute-with-specSkill
- freeze-acceptanceSkill
- independent-verifySkill
- meta-workflowSkill
- remediation-loopSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are 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.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 19
inherits that work's weight: Do not re-classify it by the literal size of the sentence.
≈ “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
Step 0 — Classify the task (the throttle)
Why this exists: the cost of process must stay below the cost of the mistake it prevents. There are two opposite failures here, and most workflows only guard against one:
- Running a full loop on a typo — process as theater.
- Straight-lining a one-line change to shared infrastructure — a real mistake shipped fast.
This step is the throttle that avoids both.
The pre-check (do this first, it's the one people skip)
Is this message starting a new task, or pushing forward / signing off an in-flight heavy one? A short reply — "yes", "do B first", "ship the second one too" — that advances heavy work inherits that work's weight. Do not re-classify it by the literal size of the sentence.
(Inheriting weight ≠ re-running the full loop for every sub-action; a sign-off or wrap-up sub-step runs at its own actual risk, but the acceptance bar stays at the original task's level.)
Then classify (only for genuinely new tasks)
| Type | Examples | Weight |
| Chat / Q&A | explain a concept, look up a value, "what is X / will Y happen" | answer directly |
| Trivial edit | typo / config / single line, and the change touches nothing shared | just do it, no self-check |
| Simple doc / code | one small function, doc wording, touching nothing shared | do it + light wrap-up check |
| Design | mechanism / architecture / skill design, selection, design docs | full loop ①→⑤, weight ①②④ |
| Build | multi-stage feature / pipeline, cross-file change, or any "fix" whose change lands on shared infra | full loop, weight ② per-stage acceptance + ④ per-stage verify |
| Exploration | the goal itself can only be set by exploring | open loop (if budgeted) |
The iron rules
- Judge by where the change lands, not by how long the message is. Anything touching shared infrastructure — hooks, schedulers, pipelines, eval, rules, automation, persistent state docs — is heavy even if it looks like "a small bug." Definition/architecture → Design; concrete feature/fix → Build; both run the full loop.
- A metric claim (a number + a metric word: recall, precision, cost, coverage, latency…) appearing → heavy by default, when this turn will actually produce a change. Pure explanation / diagnosis / questions about infra do not escalate — unless that diagnosis will drive a heavy decision or act as a quality gate.
- When unsure, go heavier.
Output
State the task class and the weight you'll run at. Then proceed: trivial → do it; design/build →
go to clarify-intent (Step ①).
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.