final-integration-check
Skill1 starsChanged 3 months agoApache-2.0View on GitHub
Verify a chief-editor-locked article after visual integration and before independent final reader review. Check final Markdown, media, captions, alt text, author and publisher metadata, and SEO fields together; validate a CMS draft payload only when that is the requested destination. Return a portable article package or draft payload with a clean cold-reader dispatch, or a precise blocker. Do not reopen editorial strategy without cause, create assets, mutate CMS, or publish.
In this repository14 other files it gives its agents.
- audit-content-librarySkill
- audit-eeatSkill
- audit-paragraph-structureSkill
- audit-tone-honestySkill
- audit-useful-actionSkill
- chief-editor-reviewSkill
- cms-draft-handoffSkill
- cold-reader-reviewSkill
- draft-articleSkill
- edit-articleSkill
- load-author-voiceSkill
- render-mermaid-infographicSkill
- run-seo-writing-workflowSkill
- visual-storytellingSkill
The file
We split this skill into the rules it gives an agent
12 rules. About 0 are common, 10 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈10
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 53
Use deliveryTarget: portable when no CMS destination has been chosen.
- Distinctive: line 61
Do not use final integration to improve the opening or reorder sections by preference.
- Very specific: line 82
Only a real source may be proof.
- Distinctive: line 108
Before CMS handoff, validate the relation payload only.
- Distinctive: line 116
Use approved inputs only:
- Distinctive: line 129
Never guess slug, canonical URL, publish date, keywords, or image IDs.
- Distinctive: line 140
Never repair another layer silently.
- Distinctive: line 210
Call a planned or missing asset integrated.
- Distinctive: line 211
Treat a generated visual as proof.
- Distinctive: line 212
Leave production markers in the CMS payload.
- Distinctive: line 215
Run the independent cold-reader review in this context or leak production inputs into its dispatch package.
- Distinctive: line 216
Read or mutate CMS, change Suggested Topic, or publish.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Final integration check
Purpose
Audit the assembled article as one reader package and prepare portable Markdown and metadata, or an exact CMS draft payload when that is the requested destination. Confirm that text, media, captions, metadata, and publication boundary preserve the chief-editor meaning lock.
This is an integration check, not another open-ended editing pass.
Responsibility
final-integration-check:
- compares final reader Markdown with the chief-editor lock;
- verifies real integrated assets against the media manifest;
- aligns claims across text, image, caption, and
alt; - checks markers, links, privacy, accessibility, and mobile behavior;
- validates trust surface and the proposed CMS payload;
- returns a
readyintegration handoff and minimal reader-surface package forcold-reader-review, or a precise owner and blocker.
It does not change the reader promise, argument, section order, or claim without cause; create a missing visual; substitute illustration for proof; read or mutate CMS; create a draft; change a Suggested Topic; or publish.
Inputs and priority
Use:
readychief-editor decision log andlockedMeaning;- locked reader Markdown and snapshot;
- Article Brief and claim permissions;
visual-storytellinghandoff with annotated draft, media manifest, production status, and deferred visuals;- real local assets or verifiable URLs;
- captions,
alt, source-of-truth, and privacy decisions; - author and publisher metadata snapshot and byline intent;
deliveryTarget: portable | flexim_draft, with portable metadata or a proposed CMS draft payload including the H1 decision;- mode:
text-only,integrated, orautomatic.
Priority: explicit user correction, chief-editor lock and Article Brief, verified source or real asset, claim permissions and product state, visual handoff, proposed payload, model assumption.
Return a changed media claim or payload to the correct owner. Never update the lock silently.
Modes
text-only: check a package without new visual assets. Returnreadyonly when the user explicitly deferred visuals or all required visual roles are fulfilled by editable Markdown or HTML artifacts.integrated: check final Markdown with every real integrated asset.automatic: ask no questions; defer optional missing visuals, block on missing required proof, and never create assets or mutate CMS.
Infer mode from production status and integrated assets when unspecified.
Use deliveryTarget: portable when no CMS destination has been chosen. A Flexim draft request uses flexim_draft; do not downgrade a blocked draft payload to portable and call the requested draft ready.
Process
1. Verify the lock
Compare title, useful action, reader, scope, product role, and section order with lockedMeaning. Compare snapshot or enumerate post-lock deltas. Allow only typo, format, and integration changes that do not alter meaning. Require a chief-editor amendment and affected reruns for a semantic delta.
Do not use final integration to improve the opening or reorder sections by preference.
2. Build final inventory
Count and record H1, H2, H3, prose paragraphs, lists and items, Markdown or HTML tables, blockquotes, code fences, reader links and destinations, media references, captions, alt, internal markers, and reader-facing key phrase placements. Keep reader scope separate from service sections and handoff.
3. Verify the media manifest
For every visual ID confirm:
- placement exists in final Markdown;
- the asset exists and uniquely matches the manifest;
- role, format, and authenticity match;
- source of truth is available;
- current and future product states are honest;
- caption carries the takeaway and qualification;
altprovides a semantic equivalent rather than duplicating the caption;- privacy and permissions pass;
- desktop and mobile variants or responsive behavior have been checked;
- filename, format, size, and path are suitable for handoff.
Only a real source may be proof. Mockups, stock, and generated illustrations never receive that status.
4. Handle deferred visuals
Classify:
required: without it, the reader promise or evidence fails;optional: it improves explanation or pace, but text stands alone;deferred_by_user: the user explicitly postponed production.
A missing required asset blocks ready. An optional or user-deferred visual does not block a text-only package when reader Markdown contains no [MEDIA: ...] marker and its production brief remains outside copy.
5. Run cross-surface claim checks
Compare every claim repeated in body copy, in-image text or UI, caption, alt, description, and SEO metadata. Number, qualification, time state, causal direction, and commercial disclosure must agree. Return any stronger unsourced visual claim to visual integration.
6. Verify reader cleanliness
Final reader Markdown must contain no [VERIFY: ...], [EVIDENCE NEEDED: ...], [MEDIA: ...], audit IDs, decision comments, service headings, target query as a production object, CMS schema or status instructions, or production-story narration without an independent reader purpose.
Do not perform the final cold-reader review inside this skill. Prepare its package from reader-visible material only: title, reader Markdown or rendered page, visible author and publisher surface, inline media, captions, alt, and reader-facing links. Exclude the Article Brief, sources, claim permissions, author profile, state, audit reports, decision log, IDs, expected verdict, and fixes. Do not reopen locked prose without an integration reason.
7. Verify trust surface
From supplied metadata check author relation and display name, intended byline, required bio or disclosure, commercial relationship with a recommended product, honest publish and update dates, and required publisher, contact, or methodology surfaces.
Before CMS handoff, validate the relation payload only. Do not claim rendered byline or page layout has been checked before a private preview exists. Keep it as a post-handoff human check.
8. Prepare the delivery package
For portable, return the actual final reader Markdown, title, description, and approved author/SEO/media metadata, preserving the existing H1. Leave unknown or destination-specific values unset. Set cmsDraftPayload: null and qa.cmsPayloadDraftOnly: null (not applicable). A missing collection, schema, slug policy, relation ID, or CMS connection does not block the portable article. Keep the selected topic's source provenance outside the reader package.
For flexim_draft, prepare the CMS draft payload below. If the user chooses Flexim later, validate its actual schema mapping in a new integration pass before mutation; a prior portable result is not proof of draft-payload readiness.
Use approved inputs only:
- collection;
- name;
status: draft;- content and
contentIncludesH1; - description;
- author, category, and tag relations;
- approved real media IDs;
- SEO component;
- fields that must remain unset;
- Suggested Topic ID and expected current status.
Never guess slug, canonical URL, publish date, keywords, or image IDs. Do not duplicate H1 or remove it without a verified presentation decision.
9. Assign blocker ownership
- semantic or contract drift →
chief_editor_review; - evidence or claim gap →
audit_eeatorresearch; - missing or wrong asset →
visual_storytellingor media production; - caption,
alt, privacy, or mobile defect →visual_storytelling; - payload or schema uncertainty → read-only preflight in
cms_draft_handoff; - preview typography or byline rendering → post-handoff
human_review.
Never repair another layer silently.
Readiness gates
Return ready only when the final package matches the meaning lock; inventory is complete and markers are absent; all required assets are real and verified or omitted under an approved text-only contract; captions, alt, privacy, and available-scope mobile behavior pass; claims align across surfaces; trust metadata is honest; the portable package is complete or the requested CMS payload is complete, draft-only, and contains no model guesses; a clean cold-reader dispatch package is complete; no publication occurred; and nextStage is cold_reader_review.
Return blocked when semantic drift, required proof, privacy, inaccessible assets, unresolved markers, or payload conflict cannot be repaired within integration scope.
Output
status: ready | blocked
mode: text-only | integrated | automatic
deliveryTarget: portable | flexim_draft
lock:
sourceSnapshot: "..."
finalSnapshot: "..."
semanticMatch: true | false
inventory:
h1: 0
h2: 0
h3: 0
paragraphs: 0
lists: 0
tables: 0
codeFences: 0
links: 0
media: 0
visuals:
productionStatus: complete | partial | deferred_by_user | not_needed
requiredReady: true | false
manifest: []
deferred: []
qa:
claimsAligned: true | false
markersRemoved: true | false
linksValid: true | false
captionsReady: true | false
altReady: true | false
privacySafe: true | false
mobileCheckedInAvailableScope: true | false
trustPayloadReady: true | false
cmsPayloadDraftOnly: true | false | null
portablePackage:
readerMarkdownRef: "..."
metadata: {}
cmsDraftPayload:
collection: "..."
fields: {}
contentIncludesH1: true | false
expectedUnsetFields: []
suggestedTopicId: null
coldReaderDispatch:
title: "reader-visible title only"
readerMarkdown: "reader-visible Markdown only"
visibleAuthorPublisher: []
mediaCaptionsAlt: []
readerLinks: []
forbiddenProductionInputsOmitted: true | false
postHandoffHumanChecks: []
warnings: []
blockers: []
nextStage: cold_reader_review | chief_editor_review | visual_storytelling | research | blockedReturn reader Markdown, media manifest, and human-readable report beside the package.
Do not
- Perform new editorial work without a semantic or integration cause.
- Call a planned or missing asset integrated.
- Treat a generated visual as proof.
- Leave production markers in the CMS payload.
- Invent metadata or publication fields.
- Claim private-preview rendering was checked before draft creation.
- Run the independent cold-reader review in this context or leak production inputs into its dispatch package.
- Read or mutate CMS, change Suggested Topic, or publish.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

