feat: good-offense planner (preemptive interventions + forward reserve deployment) #43
Merged
wasabi opened 1 week ago

Implements the "best defense is a good offense" thesis as three DRAFT/human-approved capabilities:

A. Preemptive Intervention Planner (services/threatcasting/preempt.py)

  • Read-only lead-time estimate from a flag's warning→critical band + check_frequency_hours.
  • Converts warning-band threat_flag rows into DRAFT preemptive-intervention proposals with computed lead_time_hours and severity scaled to remaining margin.
  • Writes no governed state and performs no autonomous action. CLI: python3 -m services.threatcasting preempt [--location-id UUID].

B. Forward Reserve Deployment (services/strategic_reserve/health.evaluate_trigger)

  • New preempt_threshold_pct column (migration 328) surfaces a DRAFT preempt_proposed proposal once the monitored metric enters the preempt band (preempt_pct of the way to the hard breach) — before threshold crossing.
  • Seed sets 0.8 on carbon_buffer + capability_standby reserves. Still DRAFT + human-approved; never an autonomous drawdown.

C. Proactive Advisory

  • Confirmed the existing forecast_based rule layer (spray_window / frost_risk) already provides forecast-keyed proactive advisory; no net-new code.

All proposals remain DRAFT and require human approval per governance constraints. 11 new tests in tests/test_good_offense.py (all passing).

Commits were merged into target branch
1/1
Submitter wasabi
Target main
Source feat/good-offense
Jobs
Merge Strategy
Create Merge Commit
Watchers (1)
Reference
pull request KI-43
Please wait...
Connection lost or session expired, reload to recover
Page is in error, reload to recover