docs: add cross-CLI learning control-plane implementation plan + DAG

[0xNyk]

Summary

• add executable implementation plan for LACP cross-CLI learning control-plane
• add harness task DAG aligned to existing config/harness/tasks.schema.json
• include phased 30/60/90 rollout, KPI gates, rollback trigger, and operator commands

Validation

• bin/lacp harness-validate --tasks docs/plans/2026-04-13-cross-cli-learning-control-plane.dag.json --json
  • result: ok: true, errors: 0

Notes

• planning/docs-only PR; no runtime behavior changes

[0xNyk]

Added in this push for the council consultation follow-through:

1. 7-day execution checklist

• docs/plans/2026-04-14-learning-loop-7day-checklist.md

2. Demo runbook

• docs/demos/council-next-steps-demo.md

3. Runnable demo helper script

• scripts/demos/run-council-next-steps-demo.sh

Issue breakdown created:

• Epic A: Shadow capture slice for cross-CLI learning (safe v0) #112 Epic A: Shadow capture slice
• Epic B: Advisory retrieval slice with strict caps (non-authoritative) #113 Epic B: Advisory retrieval slice
• Epic C: Guarded promotion + eval + rollback gates #114 Epic C: Guarded promotion + eval + rollback

Validation:

• Ran bash scripts/demos/run-council-next-steps-demo.sh
• Script exits cleanly in current staged state (handles not-yet-implemented learning commands gracefully).