Large Corpus Fan-Out Analysis FAQs

Question 1

Can I adjust the level of thoroughness for the analysis?

Accepted Answer

Yes, you can choose between 'Some', 'A lot', or 'Herculean' effort levels, which adjust the number of segments per worker and the amount of redundant reviews per segment.

Question 2

Does it work with source code specifically?

Accepted Answer

Yes, it includes specialized token estimation for source code (approx. 1 token per 3 characters) and can be used to map logic or find bugs across entire repositories.

Question 3

What is the purpose of the Critic stage?

Accepted Answer

Critics review the output of multiple Workers to ensure consistency, identify cross-segment patterns, and flag gaps that individual Workers might have overlooked during their narrow slice of analysis.

Question 4

How does this skill handle files that are too large for one agent?

Accepted Answer

The skill calculates segments based on token counts and distributes them across multiple Worker subagents, ensuring the total corpus is covered even if it exceeds the 200k token limit.

Question 5

What happens if a subagent hits a context limit or fails?

Accepted Answer

The skill includes failure recovery logic that can automatically split a failed subagent's workload into smaller pieces and launch new tasks to complete the work.

Large Corpus Fan-Out Analysis

主要功能

使用场景

Large Corpus Fan-Out Analysis

主要功能

使用场景