Task Decomposition FAQs

Question 1

Can I use this skill during the initial planning phase?

Accepted Answer

Yes. Pre-emptive decomposition during the planning phase is highly recommended for complex features to ensure the AI has sufficient context room for each component from the start.

Question 2

What are considered 'good' split points for a coding task?

Accepted Answer

Effective split points include boundaries between independent features, architectural layers (like Model vs. Controller), or the separation of implementation from unit testing.

Question 3

Does splitting a task cause me to lose my current progress?

Accepted Answer

No. The Decompose Task skill includes a workflow to collect partial results from active subagents and merge them into the newly created subtask branches to ensure continuity.

Question 4

How does the skill determine which tasks can run in parallel?

Accepted Answer

It performs a dependency analysis to identify tasks with no shared requirements and checks for potential file write overlaps to ensure concurrent subagents don't conflict.

Question 5

When is a task considered too large for a single context window?

Accepted Answer

A task should typically be decomposed when the token report shows usage approaching 40% of the threshold (approx. 80K tokens) or if a subagent experiences frequent context compaction events.

Task Decomposition

主要功能

使用场景

Task Decomposition

主要功能

使用场景