new(context-v, issues): two workbench gaps — gated org-create from the UI, and non-destructive entity merge (stub)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UvYzx7vDWeafnkAi2nEQeb

Author: mpstaton

context-v/issues/Merge-Organizations-Or-People-Non-Destructive-Dedupe.md

@@ -0,0 +1,67 @@
+---
+title: "Merging organizations or people — when two objects turn out to be the same entity, dedupe non-destructively"
+lede: "Sooner or later two rows ARE the same org (or person) — different imports, different spellings, one with a domain and one without. There's no merge anywhere: nothing re-points edges, unions the lists, or absorbs the loser's slug into aliases. Non-destructive but effective dedupe probably deserves its own microfrontend."
+date_created: 2026-07-24
+date_modified: 2026-07-24
+authors:
+  - Michael Staton
+augmented_with:
+  - Claude Code on Claude Fable 5
+semantic_version: 0.0.0.1
+tags:
+  - Issue
+  - Augment-It
+  - Organizations
+  - Persons
+  - Dedupe
+  - Merge
+  - Canonical-Layer
+status: Open · Stub
+---
+
+# Merge organizations or people — a stub
+
+## The need
+
+Duplicates are inevitable at capture time — the candidate gates reduce them
+but can't eliminate them (operator picks "create" when the match was real,
+two imports spell one org two ways, a person exists once with an email and
+once with a LinkedIn URL). Today discovering a duplicate is a dead end:
+nothing merges. The two rows drift apart, each accumulating links, corpus,
+streams, affiliations, and observations the other doesn't have.
+
+## Shape of the thing (jotted, not designed)
+
+- **Non-destructive.** The loser is absorbed, not deleted: slug → the
+  survivor's `aliases[]` (the `updateOrg` rename precedent), arrays union'd
+  (URL-dedup'd, the additive discipline's existing keys), observations
+  preserved with their provenance, `client_access` union'd. A tombstone or
+  redirect pointer so anything stamped with the old slug/uuid still resolves.
+- **Effective.** Affiliation edges re-pointed to the survivor (dedup'd
+  against existing edges), opportunities re-stamped (the `updateOrg` fan-out
+  precedent) — after the merge there is ONE row to find.
+- **Human-driven.** Per [[human-in-drivers-seat]]: the operator picks the
+  survivor and resolves field conflicts (which `complete_name` wins?);
+  side-by-side compare, not an auto-merge score.
+- **Perhaps its own microfrontend.** A compare-and-merge surface (two cards,
+  field-by-field pick, one confirm) doesn't fit inside the org card; a
+  dedicated remote fits the one-surface-per-job pattern the workbench
+  already follows.
+
+## Deliberate contrast to keep
+
+Opportunities are **never auto-merged** by design (duplicates across record
+sets are intentional — see the Grilling-on-DB-Resolver decisions). This merge
+is for the CANONICAL layer only: organizations and persons, where one
+real-world entity should be one row. The opportunity rule stands.
+
+## Open questions
+
+- [ ] One merge verb (`entity.merge`?) or per-table
+  (`organization.merge` / `person.merge`)?
+- [ ] Duplicate DISCOVERY — does the merge surface also hunt candidates
+  (same domain, near-name), or only act on pairs the operator brings?
+- [ ] What does the tombstone look like — a row with `merged_into`, or an
+  alias entry alone?
+- [ ] Observations: re-point subject/object to the survivor, or leave them
+  citing the absorbed row and resolve through the tombstone?

context-v/issues/Org-Workbench-Needs-Create-Organization-Behind-A-No-Match-Gate.md

@@ -0,0 +1,65 @@
+---
+title: "The Org Workbench can't create an organization — and creation must sit behind a 'be sure there is no match' gate"
+lede: "The workbench's front door only finds orgs that already exist; there's no way to mint one from the UI. When creation lands, it can't be a bare form — the operator should be shown every plausible existing match (name, slug, alias, domain) and create only past an explicit gate, same discipline as adding a person."
+date_created: 2026-07-24
+date_modified: 2026-07-24
+authors:
+  - Michael Staton
+augmented_with:
+  - Claude Code on Claude Fable 5
+semantic_version: 0.0.0.1
+tags:
+  - Issue
+  - Usability
+  - Augment-It
+  - Org-Workbench
+  - Organizations
+  - Candidate-Gate
+status: Open · Jotted
+---
+
+# Create an organization from the Org Workbench — behind the gate
+
+## The gap
+
+The Org Workbench opens with `OrgSearch` — find an existing org, load its
+card. If the org isn't in the canonical layer yet, the operator's only paths
+are indirect: the CSV/record-resolution flows, or (since today) creating one
+as a side effect of promoting a bio link
+([[Person-Bio-Pages-Are-Affiliation-Signals-Not-Just-Identity-Links]]).
+There is no first-class "create an organization" in the workbench itself.
+
+## The rule: try to be sure there is no match
+
+Creation is the dangerous half of match-or-create — a casually-minted org is
+tomorrow's duplicate. The gate discipline that already governs persons
+(`AddPersonInline`: candidates ALWAYS shown, create is an explicit choice
+past them) applies verbatim:
+
+1. Operator types a name (+ optional domain/URL).
+2. Candidates surface from every signal the resolver already scores — slug
+   (100), domain via D4 `domains[*].domain` (90), fuzzy name (60), aliases —
+   not just the autocomplete's name-contains.
+3. Zero candidates still gates: an explicit "No match — create
+   ‹name›" action, never a silent create on submit.
+4. Create seeds the thin row the existing paths seed (name → slug,
+   `client_access`, and — per today's promotion work — a domain when one was
+   given, so the new org is domain-matchable from birth).
+
+## Existing machinery (nothing new server-side, probably)
+
+- `resolveOrgRow(action:'create')` — the create verb half, already defends
+  against slug races and (since today) seeds `domains[]`.
+- `findCandidates` — the scored candidate path (`resolver.candidates`);
+  `searchOrgs` is the lighter autocomplete. The gate wants the scored one.
+- `AddAffiliationInline` — shipped today; its gate → create shape is the
+  closest UI precedent. An `OrgCreateInline` (or a mode of `OrgSearch`'s
+  empty-result state) is the org-first sibling.
+
+## Open questions
+
+- [ ] Where does it live — `OrgSearch`'s "no results" state growing a gated
+  create, or a standing ➕ beside the search box, or both?
+- [ ] Does workbench-create take a domain up front (recommended — it powers
+  the strongest match signal AND seeds the new row)?
+- [ ] Minimum viable row: name + domain + client — or also kind/notes?