Mandate Campaign Framework (Experimental / Advanced Operator)

A mandate-driven, file-backed, phase-gated campaign runtime for operators who want explicit state, explicit receipts, and bounded advancement instead of hidden workflow magic.

What is actually shipped here

• frozen-enough kernel contracts in kernel/v0.1-freeze.md
• runnable thin CLIs in tools/:
  • validate-mandate.py
  • init-campaign.py
  • advance-campaign.py
  • inspect-campaign.py
• adapter-aware runtime dispatch with explicit adapter identity
• executable reference adapter surfaces:
  • adapters/steamer/
  • adapters/openclaw-mem/
• one additional sketch adapter for genericity pressure-testing:
  • adapters/content-production/
• six seeded proof bundles with replay scripts under examples/proof-bundles/
• test coverage for runtime schema, adapter dispatch, recovery/gate semantics, and proof-bundle integrity

Honest roadmap snapshot

• Phase 0 done — kernel/runtime proof exists
• Phase 1 done — executable adapter dispatch is real
• Phase 2 done — seeded Steamer shadow proof is real
• Phase 3 done — bounded operator-loop hardening slices are real
• Phase 4 in progress — repeatability breadth is still the main gate
• Phase 5 minimally done — a second executable adapter path exists without collapsing the kernel into Steamer nouns
• Phase 6+ still open — external usability, semantics freeze, RC discipline, and stable-release posture remain ahead

That is not the same thing as a stable general orchestration product.

What Phase 3 means in practice

This repo now includes public-safe proof for narrow but real operator-loop slices:

• operator-gate escalation + replay (approve_shadow_review, reject_shadow_review)
• blocked queue-empty recovery via --resume
• cycle-budget exhaustion visibility
• stalled operator-gate timeout visibility in inspect output
• durable gate-replay and blocked-visibility close-out receipts

The scope is still intentionally tight. These are bounded operator mechanics, not a claim that every failure class is already productized.

Current closure track

The next real push is now repo-local and product-native:

1. RELEASES/phase-closure-proof-matrix.md — one compare surface for Phases 4/5/6/7/8
2. examples/operator-drills.md — public-safe drills an advanced operator can actually run
3. protocol/external-operator-usability-pack.md — the minimum pack an outside operator should be able to follow
4. protocol/semantics-freeze-checklist.md — explicit freeze-now vs still-moving semantics

This is how the remaining phases move without waiting on active Steamer, openclaw-mem, or other product-line work.

What is still not done

• broad repeatability across many domains and campaign shapes
• beginner-first onboarding
• stable compatibility promises
• broad adapter ecosystem claims
• autopilot / fully autonomous positioning

Start here

1. STATUS.md — current posture and limits
2. RELEASES/phase-closure-proof-matrix.md — remaining-phase compare surface
3. examples/operator-drills.md — public-safe drills
4. protocol/external-operator-usability-pack.md — minimum operator pack
5. protocol/semantics-freeze-checklist.md — freeze checklist
6. examples/proof-bundles/PROOF_INDEX.md — canonical replay index
7. tools/README.md — CLI/runtime surface
8. tests/ — current verifier surface

Repository map

• kernel/ — shared contracts and phase model
• adapters/ — domain bindings
• tools/ — bounded runtime CLIs
• campaigns/ — seeded file-backed campaign example
• examples/ — mandate examples, proof bundles, replay helpers, operator drills
• protocol/ — thin operator packaging, usability, and semantics docs
• RELEASES/ — release-facing proof and calibration surfaces
• tests/ — runtime and proof verification

Operator honesty

Use this repo if you want explicit state, durable receipts, and inspectable operator-gated progression. Do not use this repo if you want a polished no-code builder or magic automation claims.