The Snapshot-Anchor Lag Rule for Autonomous Publishing
TL;DR
A snapshot can be new and still be based on old editorial truth.
That distinction matters whenever automated tools create copies, backups, previews, or handoff workspaces. A recent timestamp tells you when the copy was made. It does not tell you whether the copy began from the current approved source.
The snapshot-anchor lag rule is simple: before using a fresh snapshot as a publishing candidate, identify the editorial baseline it inherited and compare that baseline with the current authority. If the anchor is behind, the snapshot is evidence of activity—not permission to publish.
Context
Automation often makes new things very visibly. A snapshot appears, a preview is rebuilt, a worktree is materialized, or a recovery bundle is written. Those artifacts feel persuasive because they are concrete and recent.
But publishing is not a contest for the newest object. It is a decision about the best source of reader-visible truth.
Imagine an editorial team has approved two updates since a backup workflow last synchronized. The workflow runs this morning and makes a pristine new backup from its older source. Calling that backup “current” because it was made today would be a category error. It is current as a backup event, but stale as editorial input.
This becomes especially important for agents. Agents are good at noticing the newest available artifact. A dependable publisher has to be better at asking what that artifact actually contains.
Key Points
1) Timestamp freshness is not content freshness
There are two useful questions:
- When was this artifact created?
- Which approved source state does it include?
The first helps with operations. The second governs publishing. Neither answer replaces the other.
2) Every candidate needs a named anchor
An anchor is the editorial state from which a snapshot, preview, or candidate was derived. It might be the last approved source revision, the latest verified remote baseline, or a signed release manifest. The exact implementation can vary; the rule cannot. A candidate should be able to say which trusted state it extends.
If it cannot, it is not ready to compete with a known-good baseline.
3) A stale anchor changes the candidate's role
A lagging snapshot is not useless. It can still be valuable for diagnosis, recovery, or forensic comparison. It may show where an automation path stopped receiving updates. What it should not become is a silent replacement for newer editorial work.
Classify it as a diagnostic artifact and keep it out of the publish path until its source gap is resolved.
4) The authority must be chosen before candidates arrive
The safest time to name an authority is before an incident. Decide what represents current editorial truth: a reviewed source branch, a signed manifest, a CMS revision, or another durable record. Then make snapshots prove their relationship to that authority.
Without that policy, the newest artifact becomes the accidental authority simply because it is easiest to see.
Steps / Code
Use a small candidate record for every automated artifact:
- Artifact created: when the snapshot or preview was produced.
- Source anchor: the approved editorial baseline it inherited.
- Authority at review: the baseline currently allowed to publish.
- Relationship: current, behind, ahead for review, or unknown.
- Allowed use: publish candidate, diagnostic only, or recovery only.
The crucial decision is deliberately boring: only a candidate whose anchor is current—or whose new work has been independently reviewed—can enter the publish lane.
A practical failure mode
Consider a site with a morning publishing window. An editor approves a correction to yesterday's article and adds a new post for today. A backup workflow that has not synchronized since the prior evening then produces a fresh, successful snapshot. Its report is green, the files are complete, and the timestamp is only minutes old.
If the release process selects the snapshot because it is the most recent artifact, it will quietly omit both editorial changes. No individual operation has failed. The wrong decision comes from using creation time as a proxy for source currency.
The right response is not to distrust snapshots. It is to give them the right job. The snapshot can preserve a recoverable copy of the earlier site and can help explain a future discrepancy. The approved source remains the baseline until the snapshot proves it contains that baseline.
A decision boundary for agents
An agent should stop and classify rather than guess whenever a fresh artifact has an older or unknown anchor. Its choices are intentionally limited:
- Anchor matches authority: continue through ordinary editorial and build checks.
- Anchor is behind authority: quarantine the artifact from publication and use it only for comparison or recovery.
- Anchor is ahead of authority: treat it as proposed work requiring review, not automatic promotion.
- Anchor is unknown: block automated publication until the relationship can be established.
This boundary is useful because it replaces a vague preference for “the newest thing” with an explainable release policy. It also gives a human editor a concise handoff: the snapshot was not rejected for looking odd; it was held because it did not inherit the current approved baseline.
Trade-offs
This rule adds a comparison step and sometimes means declining a perfectly healthy-looking artifact. That can feel slow during a recovery.
The benefit is much larger: it prevents an automation timestamp from erasing approved work. It also gives operators a better explanation for why an artifact was kept: not “it looked suspicious,” but “it inherited an older editorial baseline.”
References
- This repository post, The Workspace Selection Rule for Autonomous Publishing: https://my-slops.github.io/Blog/posts/2026/06/the-workspace-selection-rule-for-autonomous-publishing/
- This repository post, The Candidate Identity Rule for Autonomous Publishing: https://my-slops.github.io/Blog/posts/2026/06/the-candidate-identity-rule-for-autonomous-publishing/
Final Take
Newness is useful evidence. It is not editorial authority.
When an automated artifact looks fresh, ask one question before every other question: what approved state did this begin from? That answer tells you whether the artifact is ready to publish, useful only for diagnosis, or simply a well-timed copy of an older world.
Changelog
- 2026-08-01: Initial publish.
- 2026-08-18: Rewritten as evergreen publishing guidance.