Configuration

Configuration is a free-form JSON object attached to a scan. It's available to prompts as the built-in {{configuration}} variable, and it also carries a couple of engine settings.

Why it matters

Configuration is how you pass scan-wide knobs to your agents without hard-coding them into prompts. Your steps can read it and adjust behavior - scope, limits, feature flags - so one workflow adapts to different runs.

A typical starting value:

{
  "max_files": 4000,
  "include_tests": false
}

A prompt can then say "respect the limits in {{configuration}}" and the agent will see those values. open·kritt doesn't enforce a schema - the meaning of each key is whatever your workflow decides to make of it.

Engine-recognized keys

A few keys are meaningful to the engine itself:

• repeat_runs - run every concrete workflow task this many times. Repeats are sequential; after the first run, a task sees only its own earlier output for the same step and input, then returns new records to append. The accumulated task output feeds the next depth. Start with 1; use 2 or 3 when the extra model cost is worthwhile.

Note

open·kritt also stores the scan's selected post-script and skill IDs inside the configuration object for bookkeeping. Your own keys sit alongside them untouched.

Next: reserved extra.