The worker is protected by a service-token Access policy. Paste your CF-Access-Client-Id and CF-Access-Client-Secret below. They're stored in this browser's localStorage and never sent anywhere except as headers on worker requests.

Walks Comms's and Experience's Tier 1 static extractions, attempts to resolve each parentEntityKey in producedArtifacts against the current Orchestrator plan records. Surfaces three numbers per consumer: resolvable (parent exists), orphan (parent doesn't exist), unresolvable (placeholder slots in parent key — expected pre-lineage-epic). Results written to m5_ops_tier2_orphan_walk_<isoTimestamp>.

Some Tier 1 extractions are not in KV. Most agents in this build received their extractions as JSON files in chat; not all of them wrote back to KV. Upload the JSON files here as a fallback. Ops can walk against uploaded data and optionally persist it to the canonical KV keys for future use.

Click "Load registry" to read the entity hierarchy from KV.

Coverage columns (Copy / Visual) are awaiting per-instance data. Comms and Experience currently emit template catalogs, not per-tactic artifacts (see Phase 0 finding F4). Once the lineage epic lands and consumers re-extract against real produced instances, these columns populate automatically. Entity counts and tactic status below are live now.

Click "Load dashboard" to compute production status from KV.

Click "Load flow map" to render the suite's data flow from live extractions.

Click "Load backlog" to read the coordination backlog from KV. On first load, Operations seeds it from the known decisions, findings, and open work.

Click "Load queue" to compute current production state across the suite.

Click "Load posture" to read producer toggle state from KV.

⚠ Debug surface. Writes real envelopes into real KV queues. The receiving agent will process the envelope as if it came from a real upstream source. Use with intention.

Pair-coupling listener watches the producer command queues (`m5_command_comms`, `m5_command_experience`) for `asset_rejection` envelopes carrying `pairContext`. On detect, the deployment record is read, sibling assets are identified by `linkedAssetIds`, and a `pair_invalidated` envelope is emitted to each sibling's producer queue with `siblingChange` content sourced from the revised sibling's `revisionSummary` field (or a placeholder if not yet present).

Click "Scan now" to do one pass. Or enable auto-scan for continuous polling every 15s.

No cascade activity yet. Cascades trigger when an `asset_rejection` envelope is observed in a producer queue with `pairContext` populated.

To test the listener: use the Debug · envelope inject view to send an `asset_rejection` with `pairContext` set to a real deployment ID.

The smoke test picks a real deployment from a plan bundle, then drives it through the rejection → regen → re-approve → deployment-approve cycle by injecting envelopes and writing records the same shape live agents would.

What this tests: envelope contracts on the wire, handler firing (real producer code via fetchInbound→dispatch chain), state transitions in asset records, idempotency, suspension semantics, cross-agent pair-coupling cascade (Pass B real code).
What this doesn't test: Orchestrator's modal/UI logic (rejection-form behavior), generation API calls, image constraint visual adherence, operator-facing UI rendering. The harness simulates Orchestrator's UI side effects (e.g., writing the deployment verdict record) so the rest of the chain runs through real code paths.

Each run mutates real KV state. Pick a disposable test deployment, not a real one you care about. The harness doesn't undo — it leaves the trail of envelopes for inspection in Pass A.

Re-runnable. Same deployment can be re-tested by clicking "Run scenario" again; the chain runs forward from current state, not from scratch.