Diagnose, contain, recover, reconcile, and verify Azure Service Bus and Event Hubs incidents with bounded message populations, replay controls, and downstream side-effect checks.
A messaging incident is not finished when the exception stops. A Service Bus lock can expire after a payment already reached its provider. An Event Hubs checkpoint can advance before a downstream write commits. A timed-out deployment can leave two processor sets and two checkpoint prefixes in play. A broad replay can turn one outage into duplicate charges, missing projections, or an unrecoverable evidence gap.
Azure Messaging Recovery keeps the affected population visible before it recommends a mutation. It binds the namespace, entity, consumer group, partitions, application revision, SDK, incident interval, evidence time, and exact authority. Service Bus work tracks active, scheduled, deferred, in-flight, expired, dead-lettered, completed, duplicate, and unknown states. Event Hubs work records per-partition retention, positions, checkpoints, ownership, lag, and downstream commits.
The workflow covers diagnosis, reversible containment, selective dead-letter resubmission, bounded replay, ambiguous-command recovery, rollback, canaries, stop conditions, and source-to-downstream reconciliation. It keeps a quiet dashboard, zero lag, or a successful command separate from verified recovery. The bundled checker validates the canonical handoff structure and lifecycle rules. It does not query Azure, inspect message bodies, prove a provider-side effect, or authorize a change.
Version 1.0.0 was reviewed on 2026-07-18 and evaluated through Codex CLI with gpt-5.6-sol in 12 fresh read-only contexts. The candidate scored 100 across three cases, compared with 68.75 for the same agent without the skill and 72.92 for the exact free upstream. It led the upstream by 13.75 points on a duplicate-payment Service Bus incident, 20 points on an ambiguous Event Hubs checkpoint deployment, and 47.5 points on the strict canonical handoff. All three cases passed. The evaluation covers those supplied cases; it does not prove a real Azure resource, message population, checkpoint, deployment, payment record, or recovery state.
The reviewed package SHA-256 is 0492e0684227ecdbb27479bc416c62e02e6299789dfd76da254ac4d6e7f34d30. Maintenance and install support are handled by the platform-operated publisher for this listing.
Input
A concurrency change was followed by lock-loss errors, 73 dead-lettered messages, and duplicate charge reports. We can read systems and pause one consumer deployment, but we cannot change configuration, settle messages, replay, purge, deploy, or declare recovery.
Output
A blocked recovery handoff with timestamped message-state unknowns, a lock and idempotency hypothesis, reversible containment, prohibited mutations, a bounded 1-2-4 canary design, payment-effect reconciliation, stop conditions, and one read-only population action.
Input
A deployment changed the checkpoint-store prefix and timed out. Current owners and checkpoints are unknown. We can read state, pause replacement processors, and roll back, but we cannot reset checkpoints, create a consumer group, replay, redeploy, or declare recovery.
Provide the Azure namespace and entity, Service Bus subscription or Event Hubs consumer group, application and deployment revisions, SDK, incident interval, current metrics and logs, entity settings, checkpoint or message-state evidence, downstream side effects, retention and idempotency constraints, and separate read, containment, mutation, replay, rollback, approval, and recovery-declaration authority. Ask to assess, contain, recover, verify, or return the canonical JSON handoff. The skill keeps unobserved counts, actions, and outcomes explicit.
Namespace, region, tier, queue, topic and subscription, or event hub and consumer group plus the affected application, environment, SDK, and deployment or configuration revision.
Dated resource health, metrics, logs, entity settings, processor state, message or partition metadata, checkpoints, ownership, retention, error text, and last-known-good evidence. Missing items may remain explicit.
Stable message, event, business, or idempotency keys and the strongest available records for database, payment, email, inventory, or other side effects.
Separate permission to read, pause, scale, change configuration, deploy, roll back, receive, settle, resubmit, replay, change checkpoints, purge, delete, approve recovery, and declare the incident recovered.
SKILL.md; Azure Service Bus and Event Hubs incident, containment, recovery, replay, verification, reconciliation, and handoff workflow; detailed recovery framework; canonical JSON handoff contract; read-only record checker; Agent Skills interface metadata
No reviews yet.
Output
An evidence hold, read-only reconciliation of both revisions and prefixes, explicit ambiguous-state branches, rollback gates, per-partition retained replay requirements, a new operation identity, source-to-sink checks, and a blocked verdict.
Input
Return strict JSON for one production Service Bus queue. Two queue counts are observed, other message and downstream states are unknown, recovery authority is absent, and reconciliation has not run.
Output
A schema-valid canonical record with exact scope and authority, nullable message populations, evidence state, blockers, limitations, not-run verification and reconciliation, a blocked verdict, and one bounded next action.
Creator
Ssolvix6
