| name | thorough-digest |
| description | Exhaustively review local materials with parallel sub-agents so no item is skipped. USE FOR: batch process local files, digest a folder, synthesize narrative from existing materials. DO NOT USE FOR: web research, single-document summaries, slide or report writing.
|
| allowed-tools | Task, Read, Write, Edit, Bash, Glob, Grep |
Thorough Digest
Responsible for exhaustive processing and parallel synthesis of local materials; does not handle external web research.
USE FOR
- Batch processing of local files, directories, sections, or material collections
- Item-by-item extraction of key points with no items allowed to be skipped
- Building a unified narrative or comprehensive report from existing materials
DO NOT USE FOR
- Requires fetching new information from the web
- Only processing one or two documents
- The primary goal is writing slides or a formal report rather than first digesting materials
Execution Skeleton
- First inventory all input items to ensure the scope is clear.
- Follow workflow to determine grouping and parallelization strategy.
- Give each sub-agent an explicit item list, output path, and the constraint that no items may be skipped.
- Aggregate findings from each group; if needed, chain an external research skill to fill gaps.
Reference Map
Output Contract
- Default deliverables: inventory, grouping results, per-group findings, and final synthesis
- Must be able to report total input count, processed count, and coverage status
Collaboration and Handoff
- If the synthesis phase reveals critical fact gaps, hand off to an external research skill
- If materials are fully digested and the next step is writing a formal deliverable, hand off to a writing or presentation-structure skill
