Decision workbench

Build the packet before you ask the machine to sound smart.

Left side is the working packet. Right side is the run surface. This is now wired for real create actions, not just a mock.

Switching templates updates the packet type only. Use reset if you want fresh starter content.

Decision packet

Edit the shared packet contract here. The same shape should stay usable for web, CLI, and agent surfaces.

Options

Constraints

Assumptions

Stakeholders

Evidence inputs

Evidence 1

Evidence 2

Run decision

Choose depth, then generate the report.

Fast is a quicker pass. Deep spends more effort challenging the call. Both use the same packet contract and the same core runner.

Fast mode

Use when you want the first recommendation quickly.

Deep mode

Use when the call is high stakes and you want more pushback.

Expected report order:

  1. Recommendation
  2. Trust band and reasons
  3. Strongest counterarguments
  4. What would change the call
  5. Evidence summary, then expandable citations
Open demo report