All checklists

Is your scheduled publishing actually completing?

Review schedule time zones, credentials, dry runs, duplicate prevention, recovery, public results, missed runs and limits before automated web or social publishing.

Last updated

#Publishing#Automation#GitHub Actions

Official sources checked:

Did a scheduled post never appear, or did a retry publish it twice? Use GitHub Actions and publishing APIs as examples to review the complete flow. Official features are distinguished from this guide’s operational recommendations. Follow your particular platform’s documentation too. The illustrations are explanatory examples, not evidence of real publication.

CHECK 01 / 8Do schedule and publication use the same time basis?

Separate the trigger time from the deadline for an observable public result.
Separate the trigger time from the deadline for an observable public result.

Choose the operating time zone, then inspect the schedule configuration and enabled state on the default branch. GitHub Actions defaults to UTC and also supports explicit time zones. Scheduled runs can be delayed or dropped. Define when a public result is expected so a missing publication can be detected independently of the trigger.

Where to look
Schedule file · time zone · default branch · Actions enabled state
When to tick this check
The intended local time matches the configuration and an expected publication deadline is defined.

Read the official guidance ↗

CHECK 02 / 8Do credentials belong to the intended destination?

Check secret availability and actual authentication to the intended destination.
Check secret availability and actual authentication to the intended destination.

List the intended account, channel and site. Check credential availability in the actual execution environment, then use a read-only account check where supported. Record expiry and renewal. A secret name existing does not prove authentication succeeds. Keep values out of repositories, logs and AI conversations, with only the permissions needed.

Where to look
Repository or environment Secrets · account-check API · destination settings
When to tick this check
Authentication, destination, required permissions and renewal are verified without exposing values.

Read the official guidance ↗

CHECK 03 / 8Have you rehearsed without publishing externally?

Prepare the output while explicitly blocking external publishing calls.
Prepare the output while explicitly blocking external publishing calls.

Provide an explicit dry-run path that prepares the draft, images, links and destination details while blocking publication and deployment calls. Review those outputs. Manual workflow dispatch is a way to start a run; it does not itself make a run safe or simulated. This guide recommends retaining evidence that the rehearsal made no external publication.

Where to look
Manual inputs · dry-run implementation · draft artifacts and logs
When to tick this check
The complete draft is reviewed and the rehearsal is verified to avoid external publication or deployment.

Read the official guidance ↗

CHECK 04 / 8Can overlapping runs avoid duplicate publication?

Combine concurrency control with a stable job ID and publication records.
Combine concurrency control with a stable job ID and publication records.

Overlapping runs can pick the same draft. This guide recommends serializing a destination and keeping a durable job ID built from content, language and channel, with publication records. Concurrency control alone does not provide exactly-once publication. Recheck the job before sending and rehearse duplicate input to see that only one publication is prepared.

Where to look
Concurrency configuration · job ID · publication records · duplicate-input rehearsal
When to tick this check
Overlaps and repeated input are tested with durable records that prevent duplicate publication.

Read the official guidance ↗

CHECK 05 / 8Does a timeout trigger a safe recovery path?

Reconcile an uncertain destination result before choosing to resend.
Reconcile an uncertain destination result before choosing to resend.

A request can time out after the destination accepted it. This guide recommends an unknown-outcome state and querying destination results before resending. Use idempotency support when the particular API offers it; otherwise hold uncertain publications for reconciliation. Bound retries and respect service-specific Retry-After guidance. Persist completed stages so recovery can resume from the failure.

Where to look
API responses · destination lookup · saved stages · bounded retry settings
When to tick this check
Unknown and confirmed-failure outcomes are distinguished and lookup, retry and resume paths are rehearsed.

Read the official guidance ↗

CHECK 06 / 8Does a successful run match the public result?

Record draft, deployment, live verification and channel publication separately.
Record draft, deployment, live verification and channel publication separately.

Draft creation, deployment initiation, live-page verification and social publication are separate outcomes. This guide recommends retaining each destination URL or post ID and checking its title, language and image. Record partial completion when one destination fails, then resume only the unfinished one. A zero exit code alone is not proof that every channel published.

Where to look
Workflow step logs · deployment result · public URL · channel post IDs
When to tick this check
Destination results are verified and recorded as complete, failed or partially complete.

Read the official guidance ↗

CHECK 07 / 8Can you detect failures and runs that never started?

Failure alerts and independent missing-result checks cover different gaps.
Failure alerts and independent missing-result checks cover different gaps.

Test delivery of a controlled failure notification. Separately compare the last successful result and expected publication count with the deadline. A run that never starts produces no failed-job log, so this guide recommends an independent overdue-result check. Record whether work was intentionally skipped, prevented as a duplicate or genuinely missed.

Where to look
Failure notifications · last success · expected versus observed publications
When to tick this check
A test failure notification arrives and a defined check can identify an overdue missing result.

Read the official guidance ↗

CHECK 08 / 8Are retries, runtime and generation costs bounded?

Bound publication, retry and spending separately, with a recovery path.
Bound publication, retry and spending separately, with a recovery path.

Define daily publications, generation attempts, runtime and paid-API limits. Account for GitHub usage separately from external model or publishing services. This guide recommends recording attempts and usage with an alert threshold. At the limit, retain the unfinished job and notify the operator. Document who restores it and how, so the system does not silently disappear.

Where to look
Platform usage · external API billing · attempt counters · recovery procedure
When to tick this check
Actual limits are documented and a simulated limit reaches a retained pending state with notification and recovery.

Read the official guidance ↗

Copy this prompt and paste it into your AI

Replace the placeholders with your URL, tools and current situation. Copy the full text into your AI to request all eight checks, verification of the public result and any remaining account tasks.

Sources & further reading