Claim consolidation review

Read shared conventions and publication before writes.

Claims accumulate faster than pages are curated. knowledge_remember saves a claim without a subject page when no name in it matches exactly one page, and scheduled maintenance's knowledge_relink links only claims whose source page is an entity page or whose text names exactly one page. Linked claims pile up too: a hub page collects status snapshots, duplicates and claims that belong on a more specific subject. This review runs one procedure over a chosen scope. Work in this order: choose the scope, link, then duplicates, replaced claims and snapshot series, then page prose.

1. Choose the scope and enumerate it

Call knowledge_status for the intended scope and knowledge_ontology for the current fingerprint. Then enumerate exactly one starting set:

• Unlinked residue (the default, and the only scope of scribe runs). Run knowledge_relink without dryRun first so every remembered claim it can link is linked. Then list the residue with knowledge_relink dryRun: true, include: "all" and limit: 10, following nextCursor: every current claim with no linked page and no source page, whichever tool saved it. Each call reports at most 10 examples per outcome, so a limit of 10 returns every claim it examined.
• One page, for a hub page or a page over its size budget. List its current claims with knowledge_connections id, direction: "claims", view: "full" and includeCards: true, following nextOffset with the same filters. Restart when snapshot changes. On a hub, group from the cards first and work one group at a time; do not read every claim in full up front.
• One subject, optionally since a date. When the subject has a page, list its claims with knowledge_timeline entity, timeField: "recorded_at" and since, following nextOffset. Also search for claims filed elsewhere or unlinked with knowledge_grep kind: "claim", the subject's distinctive names (a PR number, a commit SHA, a feature name) and the same since/timeField. Keep only claims that are actually about the subject.

State the scope in the report. Read each claim you will change with knowledge_record and metadataView: "full". Its subject_hint, when present, is the saver's own description of what the claim is about and the strongest clue to its subject. Group the claims by what they are about (one feature, PR, commit, decision thread, person or preference) before choosing pages. Most claims make sense only next to their neighbours: a series of status snapshots about one PR usually ends in an outcome claim.

2. Choose subject pages

For each group, find candidate pages with knowledge_grep and knowledge_lookup, and read their State and Timeline. Ignore withdrawn pages and fixture pages.

• The primary subject is what the claim asserts something about; list it first in entities. Add a secondary subject only when the claim also states something about that entity, such as the person who made the decision or the company whose product was configured. A passing mention is not a subject. Two or three subjects is usually enough.
• The primary subject is the most specific page the claim is about: a claim about one PR belongs on that PR's page, not only on the project the PR belongs to. The broader page stays a subject only when the claim says something about it, such as a project-level decision, policy or release.
• A preference or decision stated by a person belongs on that person's page. Do not turn one person's preference into a project or company policy.
• Match identity on evidence, not on similar names. When a claim names something that might be an existing page (for example a product name that could be a different service), confirm it from evidence such as configuration, a URL or the page's own aliases. If it stays uncertain, leave the claim where it is and report it.

Link with knowledge_annotate entities, passing the complete list (it replaces the existing one) and the claim's current revision and hash. Assertion text, evidence and lifecycle stay unchanged. Link a claim before you supersede it: superseded claims cannot be relinked.

3. Create missing pages and relink to them

Linking every claim to the nearest existing page can bury a real subject in a broad page's Timeline. Create a page when:

• a discrete subject (a feature, project, tool or substantial PR) has roughly five or more claims, or has its own design, proof and outcome history; or
• a person or organisation authored, owns or decided something in the claims, has no page, and has identity evidence (for example the PR author shown by gh pr view).

Do not create a page for a single mention or a one-off fix that a Timeline entry covers. Use knowledge_save_page with the ontology type, aliases, and the identity evidence as source. Person pages use identity.status: partial unless the identity is fully established. Write a short State that cites the claims and a dated Timeline. Link the new page from its parent page.

Then relink the group's claims, including the snapshots step 4 will supersede. The new page becomes each claim's primary subject. Keep the broad page the claims were filed on (often the hub this review started from) only for claims that say something about it, by the rule in step 2; drop it from the rest. Keep other existing secondary subjects that still qualify. Under the threshold, leave the claims on their current pages.

4. Retire duplicates, replaced claims and snapshot series

• Duplicate: two claims assert the same fact, decision or preference about the same event. Keep the one with owner-approved recall, otherwise the earlier one, otherwise the one with fuller attribution. Copy any evidence the survivor lacks with knowledge_annotate addEvidence, then knowledge_withdraw the other with a reason that names the survivor's ID.
• Replaced: a later claim explicitly corrects the earlier one (for example an assistant correcting its own report) or decides a question the earlier claim left open. Supersede the earlier claim by the later one with knowledge_supersede, with a reason that says what replaced it. Do not use knowledge_correct here: it publishes a new copy of the replacement, which creates another duplicate.
• Snapshot series: several status snapshots about one specific item (the same PR, commit or decision thread: CI red at one head, review requested, CI green) followed by an outcome claim about the same item (merged, closed, reverted, released, decided, abandoned). Supersede every snapshot in the series by the outcome in one knowledge_supersede call (up to 25 claims). All of these must hold, from the claims' own text and dates:
  • every snapshot and the outcome name the same specific item, not merely the same project;
  • the outcome is dated after every snapshot it supersedes and states a final state, not another intermediate one;
  • the outcome and the snapshots share a subject page (link them first, step 3); the tool rejects claims with no common page. A snapshot dated after the outcome, or one about a different head or a different item, stays.
• Neither: snapshots of different moments that no later outcome replaces (CI red, then CI green, with no merge yet) are history, not errors. Keep them. Contradiction alone does not justify a withdrawal or a supersession. When unsure, keep both and report the pair.

knowledge_supersede keeps every superseded claim's text, evidence, provenance and history, records superseded_by and the reason on it, and leaves the superseding claim unchanged. Default reads stop showing superseded claims; historical reads (historical: true, lifecycle: "all") and knowledge_history still do. If a supersession was wrong, call it again with undo: true, the same by claim and the fences from its receipt's items.

5. Update page prose

Linking makes the claims reachable from their pages; it does not change what a page says. Edit a page only when the claims change its durable state: a decision, a preference, a landed or released change, a new capability, a resolved open question, or page text that later claims show to be wrong.

• Collapse a series of progress snapshots into its outcome. Snapshots (CI runs, open PRs, local commits) are at most dated Timeline entries, never State.
• Check mutable status live before writing it as current (for example gh pr view for merge state) and say when you checked. A merge is not a deployment, and a deployment is not proof that a given Gateway runs the code.
• When a design or decision page contradicts a later decision, add a dated amendment next to the original text rather than rewriting it. Mark answered open questions as resolved and keep the original question.
• Cite every added sentence with [[claim-id|label]] links to the claims.
• Page bodies have a size budget (page budgets). A state: "rejected" edit saved nothing, but the linked claims already hold the facts. Move a section to its subject's page with knowledge_extract_section (creating the page per step 3 when needed), reread, and retry. Never shorten other prose to make room. Extraction makes the destination the first subject of the claims the section cites and keeps their existing subjects, including the page you extracted from. Apply step 2 to each of those claims and drop the source page with knowledge_annotate entities where the claim says nothing about it.

Use knowledge_edit_body for section edits. Compute UTF-16 offsets from a fresh full read, and reread the revision and hash before every edit: other sessions edit busy pages concurrently. Prefer inserting text over replacing it. Replace a whole State section only when all of it is stale, and keep the claim links it already had.

6. Verify and report

Run knowledge_verify with expectedEntities for each linked claim and expectedLinks for each new or edited page. Enumerate the scope again: new claims may have arrived while you worked. Handle them the same way or report them.

Report the scope, the number of claims examined, linked and relinked, pages created, claims superseded (with the superseding claim for each series), claims withdrawn (with the surviving claim for each), pages edited, and claims left unchanged that you considered, with the reason for each. When running unattended, never ask for a decision and never guess: leave uncertain identities, uncertain duplicates, snapshot series whose outcome is not certain, and pages you would only create on a guess untouched, and list them in the report for a person.

Scribe runs

The Knowledge Consolidation scribe runs this review unattended over the unlinked residue, on the schedule in its prompt. Its prompt checks access and sets up its automation; this section is only the work of one run.

• Run step 1 for the unlinked residue. When the residue is empty, stop there and report zero claims examined.
• Otherwise work through steps 2 to 6, and check the residue again before stopping, since claims saved during the run are part of the same backlog.
• Do not ask questions. Give the step 6 report; the closing status line counts claims examined, linked, superseded, withdrawn and left.