Forge jobs
Inspect queued and scheduled work for durability, idempotency, retries, poison messages, and operability. Activate automatically for queues, workers, cron, scheduled functions, and outbox consumers when that concern is relevant to a software-engineering request.From its SKILL.md
npx -y skills add is-bo/fullstack-forge-skill --skill forge-jobsAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
4.4 KB, 841 tokens by cl100k_base, as published. Nobody here has run it
forge-jobs: Background jobs
Purpose
Inspect queued and scheduled work for durability, idempotency, retries, poison messages, and operability.
This is an agent playbook, not a claim of standalone analyzer coverage. Apply
fullstack-forge/references/shared/module-contract.md
for common applicability, evidence, command-safety, mutation, verification, and completion rules.
Never hide failed checks or claim that an operation ran when it did not.
Automatic activation signals
Activate when a request or direct repository evidence involves background jobs, when
the user explicitly names forge-jobs, or when discovery proves an applicable boundary.
- Queues, workers, cron, scheduled functions, and outbox consumers
When not to activate
- Applications with no asynchronous or scheduled execution
Automated support
Relevant discovery inputs are:
- queue and scheduler configuration
- producers and consumers
- job tests
Available deterministic support, where present:
- Use
inspect-routesfor its bounded evidence when present; treat unavailable runtime evidence asNOT_VERIFIED.
Agent inspection procedure
- Inventory queues, workers, and scheduled jobs with their triggers, and trace one job from enqueue to completion including the failure path.
- Verify idempotency: determine what happens when the same job runs twice, concurrently, or after a partial failure.
- Inspect retry policy, backoff, limits, and the destination of permanently failing jobs (dead-letter or silent loss).
- Check transaction boundaries: whether enqueue happens atomically with the state change that requires it (outbox or equivalent).
- Verify observability: how an operator would find a stuck queue, a poison message, or a job that never ran, and how deploys interact with in-flight jobs.
Manual inspection requirements:
- Review operational replay and poison-message runbooks
- Confirm provider retention and retry settings outside the repository
Stack-specific guidance:
- Use provider-native acknowledgement and visibility-timeout semantics
Evidence to collect
For formal findings, also follow fullstack-forge/references/PROTOCOL.md. Record the module's
inspected boundary, relevant tests, direct observations, and unavailable evidence.
Primary standards used as criteria, not proof of compliance:
- CloudEvents concepts
- NIST SSDF
Common production failures
- Trace enqueue-to-effect flow and delivery guarantees
- Inspect idempotency keys, leases, timeouts, retries, jitter, dead-letter handling, ordering, concurrency, and payload versioning
- Verify transaction boundaries, outbox/inbox behavior, observability, replay, and shutdown
Missing-control checks
For every applicable criterion below, attach direct evidence or record a reasoned
NOT_APPLICABLE, NOT_VERIFIED, or BLOCKED status. The list is a routing checklist, not
evidence by itself.
- Job idempotency
- Duplicate execution
- Retry limits
- Exponential backoff
- Dead-letter handling
- Timeouts
- Scheduling
- Concurrency
- Ordering
- Partial failures
- Transaction boundaries
- Poison messages
- Cancellation
- Progress reporting
- Safe deployment while jobs run
- Job observability
- Queue saturation
- Backpressure
- Email, invoice, image, reconciliation, report, notification, and import jobs
Commands and tools
- Run
forge jobs audit --jsonorfullstack-forge jobs audit --jsonwhen an explicit audit is requested and the CLI is installed. Normal feature work does not require it. - Use the deterministic support named above only for its documented bounded evidence.
Safe fixes
- Add bounded retry metadata, structured job logging, and payload validation
- Add tests for duplicate delivery
Approval-required changes
- Introducing a queue or changing delivery and ordering semantics
Verification
- Deliver the same job twice and observe one durable outcome
- Force failure through retry and dead-letter paths
Completion contract
Apply the shared module contract and the module-specific limitations below.
Known limitations
- Managed queue policy is NOT_VERIFIED without configuration output
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.