Clarify, then delegate
The main Agent defines the goal, source materials and delivery requirements, then delegates independent work to specialist sub-Agents. It identifies gaps and assembles returned work; sub-Agents execute within authorized boundaries and return findings and open questions.
- Input
- Statements and supporting records
- Returns
- Key fields and differences
- Input
- Contracts and legal records
- Returns
- Clause questions and sources
- Input
- Business and project records
- Returns
- Questions and confirmation items
Keep execution visible
Each subtask keeps its messages and execution process in an independent context. Progress, issues and results return to the main task, while concurrency, collaboration depth and timeouts remain policy controlled.
- Run independent work in parallel
- Tasks without dependencies can progress together.
- Create pending items for people
- Missing facts or approvals go to the responsible colleague.
Assemble results
The main Agent assembles sources, differences and confirmation items into a reviewable deliverable with a source index, cross-check tables and a list of discrepancies. Items requiring professional judgment return to the original task after review.
- Keep supporting evidence
- Source indexes, cross-checks and differences stay with the deliverable.
- Connect human judgment
- Items return to the original task after professional review.
Project due diligence working papers
- Source index
- Files, versions and source locations
- Cross-check table
- Differences across financial, legal and business fields
- Differences and issues
- Missing, conflicting and judgment-dependent issues
