Skill · Prompt

Split the work into focused slices, each with clean context

A prompt that makes the agent break a big task into a few independent, checkable pieces — instead of cramming it all into one blind run.

TypePrompt

LanguageĐọc bản tiếng Việt →

You hand over a big, multi-part task in one long prompt. The agent takes it all on, runs straight through, and returns one large block — and now you have to read the whole thing to find where it went wrong, or which part dragged another off course. The context got stuffed with everything at once, so each part came out blander than it would have alone.

A big task is rarely one task — it's a few tasks stuck together. Break it into pieces with their own goals and each one needs only its slice of context, and you get to check each piece before it can drag the whole block off course. This is orchestration even when you only have one agent: split it to see it clearly, not to run it faster.

✕ One block

Cram the whole task into one long prompt

▼

Context diluted, later parts blander than the first

▼

Where's it wrong? You read the whole block to find out

✓ A few clean slices

2-4 slices, each its own goal + clean context

▼

Run one at a time, check each piece

▼

Reconcile + flag where the slices conflict

Paste it when the task is big enough that you can feel "it's made of a few parts." The trick is making it state the split first and wait for your nod — because the likeliest failure isn't inside any one slice, it's at the seams between them: what belongs to which, and whether one slice needs another's result.

Paste into the chat

Before doing this, split it into focused slices I can check independently:

1. Break the task into 2-4 sub-tasks, each with a clear, separate goal.
2. For each slice, list ONLY the context it actually needs — don't carry the whole task into every piece.
3. Do them one at a time (or hand each to a fresh sub-agent), keeping each slice's context clean.
4. At the end, reconcile: show how the pieces fit together and flag any conflicts between them.

Tell me the split first — the slices and what context each gets — and wait for my OK before you start.

Prefer your agent install it?

Add this rule to my CLAUDE.md, verbatim:

Before doing this, split it into focused slices I can check independently:

1. Break the task into 2-4 sub-tasks, each with a clear, separate goal.
2. For each slice, list ONLY the context it actually needs — don't carry the whole task into every piece.
3. Do them one at a time (or hand each to a fresh sub-agent), keeping each slice's context clean.
4. At the end, reconcile: show how the pieces fit together and flag any conflicts between them.

Tell me the split first — the slices and what context each gets — and wait for my OK before you start.