## What & why
The audit-log viewer lived as a full-screen page reachable only via a
"View Logs" button buried in the **Security** tab. This surfaces it as
the **third tab in General settings** (`General | Security | Logs`),
consistent with the other tabs.
## Changes
- **Relocated** the event-logs module
`pages/settings/security/event-logs/` → `modules/settings/event-logs/`
and render it as tab content instead of a `FullScreenContainer` page.
Dropped `SettingsPath.EventLogs`, its route, and the fullscreen handling
in favor of the `general#logs` hash tab.
- **Security tab:** removed the "View Logs" entry; kept the
log-retention setting there.
- **In-tab gating** (shown to users with the Security permission):
Enterprise upgrade card when not entitled, a clear "ClickHouse not
configured" placeholder otherwise (derived from client config), and the
query is skipped when disabled. Replaces a bespoke error component that
string-matched error messages with the shared `SettingsEmptyPlaceholder`
/ `SettingsEnterpriseFeatureGateCard`.
- **Layout:** boxed content column with the table selector + filters
grouped in a `Card` and the results table below, matching settings
conventions. Kept the existing fixed filters (page/event name, member,
period) rather than recreating the record-view filter chips (those are
tightly coupled to record/view context).
Frontend + `twenty-shared` only — no changes to the log query or data.
## Test plan
- [x] `npx nx typecheck twenty-front` and `npx nx lint twenty-front`
pass
- [x] Settings → General shows three tabs; Logs is the third; breadcrumb
stays "Workspace / General"
- [x] With Enterprise + ClickHouse: table selector, filters, refresh,
and the paginated table work
- [x] Non-Enterprise: Enterprise upgrade card shown; no failing query
fires
- [ ] Enterprise without ClickHouse: shows the "ClickHouse not
configured" placeholder
- [ ] Security tab still shows the log-retention setting and the "View
Logs" button is gone
- [ ] A user without the Security permission sees neither the Security
nor Logs tab
## Summary
- Adds `@argos-ci/storybook` vitest plugin to `twenty-ui` for automatic
screenshot capture during vitest storybook tests
- Uploads captured screenshots (PNG, ~5MB) as a CI artifact instead of
passing the full storybook build
- Updates the visual regression dispatch workflow to pass
`mode=argos-screenshots` to ci-privileged, which then uploads
screenshots to Argos via CLI
This replaces the 10-minute Storybook screenshot capture with a ~30s
vitest browser-mode approach. The heavy screenshot work happens on free
public runners, while ci-privileged only handles the Argos API upload
(keeping secrets private).
## Architecture
```
twenty (public, free runners) ci-privileged (private)
───────────────────────────── ────────────────────────
1. Build storybook-static 4. Download screenshots artifact
2. Vitest captures screenshots 5. `argos upload` → Argos API
3. Upload screenshots artifact 6. Poll for results
7. Post PR comment
```
## Test plan
- [x] Verified locally: vitest captures 225 screenshots in ~28s
- [x] Verified `@argos-ci/cli upload` successfully creates Argos build
from captured screenshots
- [x] Argos diffs computed and results visible via API
- [ ] CI runs end-to-end on a PR
- **Add delete many**, `delete_many_{object}` added alongside the
existing `delete_one_{object}`.
- **Uniformize naming**, crud module, type names, and MCP helper
constants renamed for consistency.
- **Optimize tool schema (learn phase)**
- `find_many(_companies)`: **7 158 → 2 700 tokens**
- `find_one(_company)`: **280 → 126 tokens**
- ....
- Main mechanism: `reused: 'ref'` (line 7 of
`to-tool-json-schema.util.ts`). Zod walks the schema tree, tracks which
Zod schema instances appear more than once, and emits each reused
instance exactly once in `$defs`, replacing all subsequent occurrences
with a `$ref`. Works because filter and value schemas are now extracted
as shared objects.
- **Optimize system prompt (tool catalog)**, DATABASE_CRUD section
restructured to list operation patterns (`find_many_{object}`, …) once +
objects once, instead of the full N×M cross-product of tool names.
- **Optimize execute_tool**, shared record-properties schema (same
`$defs` deduplication applies at call time); introduced `upsert_many`;
added `selectedFields` to `find_*` so the agent only fetches the fields
it needs.
Fixes: #21031
### Root cause:
This feature was never fully implemented even though the placeholder
text suggested it was supported. It could only update one handle at a
time.
### Fix
Fixed the zod validation to validate each handle separately.
Used `useCreateManyRecords` to update multiple handles at the same time.
### Before:
<img width="2032" height="1162" alt="Screenshot 2026-05-29 at 3 25
12 PM"
src="https://github.com/user-attachments/assets/ae6b6ae3-ed38-4410-801e-11f514773681"
/>
### After:
https://github.com/user-attachments/assets/69354129-422a-41de-baf7-fa5a28f01f3f
---------
Co-authored-by: cubic-dev-ai[bot] <191113872+cubic-dev-ai[bot]@users.noreply.github.com>
Co-authored-by: Etienne <[email protected]>
Workspace-aware initialize.instructions
- Deleted the static mcp-server-instructions.const.ts
- Created build-mcp-server-instructions.util.ts — a comprehensive system
prompt with identity, object list, tool grammar, routing decision tree,
intent mapping, skills vs tools, safety constraints, and data efficiency
guidelines
- Created McpInstructionBuilderService — fetches workspace-specific
object names + skill names and injects them into the instructions
Hide/deprecate get_tool_catalog
Benefit : skip first MCP call (tools are included in instruction)
Surfaces per-step "Logs" tabs in the workflow run side panel so users
can see what each step actually did (model + tokens + tool calls for AI,
console output for serverless functions, request/response for HTTP,
recipients/body for Email).
<img width="546" height="501" alt="ai_agent_without_websearch"
src="https://github.com/user-attachments/assets/c6ca3518-9489-4484-a570-3d0569ff3b03"
/>
## Storage
- New `stepLogs` JSONB column on the `workflowRun` workspace entity,
typed as `Record<string, WorkflowRunStepLog>` (keyed by step id).
- Schema lives in `twenty-shared`: `workflowRunStepLogSchema` with a
discriminated `details.type` union for `AI_AGENT | CODE | HTTP_REQUEST |
EMAIL` — frontends and backends consume the same Zod-inferred type.
- Field is added to existing workspaces via a workspace upgrade command
(`2-9 add-workflow-run-step-logs-field`); the standard-object metadata
declares it for new workspaces.
- Writes happen atomically per step in
`WorkflowRunStepLogWorkspaceService.setStepLog` using `jsonb_set`. That
lets concurrent steps in the same run write their own keys without
contending with the existing lock around `workflowRun.state`.
- Per-step payload is hard-capped at 256 KB; anything larger is dropped
with a `logger.warn`, so a pathological tool call can never bloat a row.
See below for more information.
## How logs are produced
**Aalmost everything was already being collected; this PR mostly
persists and renders it.**
- **AI agent** — `AgentAsyncExecutorService` already tracked token
usage, model id, native web-search count, and the AI SDK's `steps[]`. We
map those into the log via `mapAiStepsToToolCallLogs` (`searchVector`
stripped from record outputs, per-call input/output capped at 32/64 KB,
max 200 tool calls per step). The only new measurement is a wall-clock
`durationMs` taken around `executeAgent`, and we now fold native
web-search cost into the displayed `totalCostInDollars` (it was already
billed, just not shown).
- **Code / serverless function** — reuses the `console.log` output the
function runner already returns (`logsByLevel`);
`build-code-step-log.util` only repackages it.
- **HTTP request** — built from the action's existing input/output via
`build-http-request-step-log.util`. No new signals collected.
- **Email (send / draft)** — added `sanitizedHtmlBody` + `plainTextBody`
to the existing tool outputs (a small additive change), then
`build-email-step-log.util` consumes them.
No additional AI inference or external calls are made for logging — the
cost is a small CPU overhead per step plus the JSONB write.
## Security
The log surface intentionally shows whatever the workflow touched, which
made redaction and sanitization the main design concern.
- **HTTP — secrets in headers**: existing `SENSITIVE_HEADER_NAMES` set
(Authorization, Cookie, …) replaced with `[redacted]` in both request
and response.
- **HTTP — secrets in URLs**: `SENSITIVE_URL_PARAM_NAMES` (e.g.
`api_key`, `token`, `access_token`) replaced in the query string via
`URL`-based parsing.
- **HTTP — secrets in bodies**: `SENSITIVE_BODY_KEY_REGEX` deep-walks
JSON request/response bodies (object input or stringified JSON) and
redacts matching keys. Applied to the `error` field too, since
transport-layer errors sometimes embed structured payloads.
- **Email — XSS risk in body preview**: tool outputs now expose a
server-side `sanitizedHtmlBody`; the log builder prefers it over the raw
user-authored `input.body`, with `plainTextBody` as a second fallback.
The original raw body is only used if sanitization didn't happen (e.g.
tool failed before composing).
- **AI — internal/noisy data**: `searchVector` (Postgres tsvector
strings) is stripped from record outputs returned by Twenty tools to
avoid leaking internal full-text-search payloads.
- **DB bloat / runaway agents**: 256 KB per-step cap + 32 KB / 64 KB
per-tool-call input/output cap + 200 tool calls per step.
<img width="547" height="307" alt="logic_function"
src="https://github.com/user-attachments/assets/dd4a3d16-67f2-434b-95b3-bdcaf9ed053d"
/>
## More details on Log size & truncation
Logs are stored in `workflowRun.stepLogs` (JSONB), keyed by `stepId`.
### Per-step cap
Each step's log is hard-capped at **256 KB** (`MAX_STEP_LOG_BYTES` in
`WorkflowRunStepLogWorkspaceService.setStepLog`).
For ~99% of workflows this is roomy — typical real-world sizes:
- Code / serverless function: 1–20 KB
- HTTP request: 5–70 KB
- Email: 5–30 KB
- AI agent (a handful of tool calls): 5–50 KB
### Two layers of bounding
1. **Per-field truncation** in each builder (before writing):
- **Code**: ≤ 500 entries, ≤ 4 KB per message, ≤ 8 KB stack trace
- **HTTP**: ≤ 32 KB per body (request + response), UTF-8 byte-aware
- **Email**: ≤ 8 KB body preview, UTF-8 byte-aware
- **AI agent**: ≤ 32 KB tool input, ≤ 64 KB tool output, ≤ 200 tool
calls/step
2. **Global per-step safety net** at write time: if the assembled
`stepLog` still exceeds 256 KB, the write is **dropped entirely** with a
`logger.warn`. The workflow itself keeps running unaffected.
### What this means in practice
- **Safe**: workflow execution, step results, downstream steps — never
blocked by log size.
- **Safe**: iterators (each iteration overwrites the previous log for
that `stepId`, so they can't accumulate).
- **Safe**: step retries (same `stepId` is overwritten, not appended).
- **Possible**: an AI agent step with many large tool outputs (e.g., 50+
heavy `web_search` calls) can exceed 256 KB → the **entire** step's log
is dropped, side panel shows "No logs were recorded for this step". The
user has no explicit signal that the log was dropped due to size (only
server-side warn).
- **Possible** (theoretical): a workflow with hundreds of distinct steps
could push the row toward Postgres's internal ~256 MB jsonb limit.
Beyond that, individual `jsonb_set` writes would error and be swallowed
by the action's try/catch — workflow still completes.
### Possible future hardening (not in this PR)
- Replace "drop entire log" with a stub that preserves the summary card
(cost, duration, status) and marks `truncated.reason = 'size_cap'`.
- Surface size-drops in the UI (similar to the existing
`<StyledTruncatedNotice>`).
- Emit a metric so dropped logs are observable in dashboards.
learn_tools(["find_many_companies"]) goes from ~170ms → ~2ms (85x
faster, measured with 40 objects / empty fields — real workspaces would
see even bigger savings). Previously it generated schemas for all ~250
tools and discarded 249;
now it generates exactly the requested one(s).
## Before
https://github.com/user-attachments/assets/5108a9d8-2017-41d5-855c-98714cbd4237
## After
https://github.com/user-attachments/assets/0a78d1e1-354f-4f3f-8ec4-6f46517619e4
## Summary
- Fixes visual flickering/glitching in the workflow show page header and
canvas when editing an active workflow or discarding a draft
- Root cause: SSE events re-added discarded drafts to Apollo cache, and
multiple hook instances had independent state causing version
oscillation between DRAFT and ACTIVE
- Rewrites `useWorkflowWithCurrentVersion` with a module-level
`discardedDraftId` variable shared across all instances, Apollo cache
seeding in mutation callbacks, and `lastValidResult` caching to prevent
null renders
## Test plan
- [x] Open a workflow show page with an ACTIVE workflow
- [x] Drag a node to change position → verify no flicker, status shows
DRAFT smoothly
- [x] Discard the draft → verify header does NOT flicker between
DRAFT/ACTIVE, position resets cleanly
- [x] Click on manual trigger and edit settings → verify the edit works
(draft created, settings saved)
- [x] Repeat discard + edit cycle multiple times to confirm stability
Fixes https://github.com/twentyhq/twenty/issues/21043
## Context
Newly created fields were never added to FIELDS widgets, regardless of
the "Set fields created in the future as visible" toggle. The widget's
newFieldDefaultVisibility was null on widgets that never explicitly set
it (it was never populated at creation), so the backend skipped them and
no view field was created.
## Implementation
Keep newFieldDefaultVisibility nullable with false (not visible) as the
behavior when not provided.
The FE now reflects that properly and shows "un-toggled" when it's null
(iso with BE behavior).
The fix also ensures the value is explicitly set to true wherever it
should be:
- Set newFieldDefaultVisibility: true at every FIELDS widget creation
path (backend default record-page layout, frontend
createDefaultFieldsWidget + useTemporaryFieldsConfiguration);
- Added a 2-9 workspace upgrade command that backfills true onto
existing standard FIELDS widgets where the value is null.
Closes#18928
## Problem
When a JWT access token expires while the AI chat is streaming a
response, the SSE connection drops and `graphql-sse` calls the retry
callback. The previous implementation would wait, then destroy the SSE
client but never refreshed the token. On the next connection attempt the
client reused the same expired token, eventually triggering an
`UNAUTHENTICATED` error that redirected the user to the login screen.
## Solution
Add proactive token renewal inside `useHandleSseClientConnectionRetry`
before each reconnect attempt:
- Uses a module-level `let renewalPromise` variable to deduplicate
concurrent renewal requests , the exactpattern used in
`ApolloFactory.ts`
- Calls `renewToken` via `retryWithBackoff` against the `/metadata`
endpoint
- Writes the fresh token pair into the Jotai store ,the SSE client's
`headers()` callback picks it up automatically on reconnect
- If renewal fails -> falls back to destroying the SSE client as before
## Files changed
-
`packages/twenty-front/src/modules/sse-db-event/hooks/useHandleSseClientConnectionRetry.ts`
## Notes
This addresses the two issues from the previous review:
- No `useRef` using module-level variable instead
- CI passing removed the `CombinedGraphQLErrors` import
---------
Co-authored-by: Etienne <[email protected]>
# Introduction
Gate what connected account can be ingested in case of ai mcp user
workspace agnostic funnel to only the workspace shared connected account
Added a quick win intregration tests on seeded connected accounts ( that
wasn't covered but already protected fix impacts only the mcp )
Refactored the API slightly too
## Notice
This mean there's a breaking change in the product behavior
Whereas before a non user workspace related mcp interaction would might
have fallback on any private user connected account it will now only
search for workspace visible listed ones
## Summary
- Commit b330105470 removed `dotenv` and `zod` from
`packages/twenty-sdk/package.json` dependencies but did not run `yarn
install`, leaving the lockfile out of sync.
- `yarn install --immutable` in CI was failing with `YN0028: The
lockfile would have been modified by this install` on those two
packages.
- This PR just runs `yarn install` to sync the lockfile — no logic
changes.
## Root cause
[Failing CI
run](https://github.com/twentyhq/twenty-infra/actions/runs/26883193920/job/79288298420):
```
YN0028: - dotenv: "npm:^16.4.0"
YN0028: - zod: "npm:^4.1.11"
YN0028: The lockfile would have been modified by this install, which is explicitly forbidden.
```
## Changes
`yarn.lock` only — removes the `dotenv@^16.4.0` range and the
`twenty-sdk` metadata entries for `dotenv` and `zod`.
Automated daily sync of `ai-providers.json` from
[models.dev](https://models.dev).
This PR updates pricing, context windows, and model availability based
on the latest data.
New models meeting inclusion criteria (tool calling, pricing data,
context limits) are added automatically.
Deprecated models are detected based on cost-efficiency within the same
model family.
**Please review before merging** — verify no critical models were
incorrectly deprecated.
Co-authored-by: FelixMalfait <[email protected]>
## Summary
Partner marketplace UI fixes and CTA improvements on the website
(`twenty-website`).
## Changes
- **Find a partner buttons** → link to `/partners/list` (hero + signoff
sections) instead of opening the contact modal.
- **Chip row layout fix** → align partner chip rows to the first chip's
baseline. In the narrow 3-column grid, category chips wrap to multiple
lines; centering floated the label to the middle and visually lifted the
first chip into the row above (e.g. "Custom Development" appeared under
Languages). Baseline keeps the label pinned beside the first chip.
- **Card CTA** → replace the conditional "Book a call" calendar CTA with
a "View profile" link to the partner profile page, shown on every card
regardless of whether a calendar link exists. Keeps card heights
consistent across the grid.
- **Card avatar** → show the partner's real profile picture when a safe
`http(s)` URL is present, keeping the initials block as the fallback.
- **Profile "Contact <partner>" CTA** → when a partner has no booking
(calendar) link, show a "Contact <partner>" button (alongside LinkedIn
if present) that opens a `mailto:[email protected]` with a partner-named
subject and a pre-filled body prompt.
## Tests
- New: `PartnerAvatar`, `PartnerProfileCtas` unit tests.
- Updated: `PartnerCard` tests for the View-profile CTA.
- All `Partner*` website tests pass.
## Screenshots
<img width="1440" height="816" alt="Screenshot 2026-06-02 at 23 40 55"
src="https://github.com/user-attachments/assets/bd88a9be-3297-4d7f-891c-c9d403d2b4d9"
/>
<img width="1441" height="818" alt="Screenshot 2026-06-02 at 23 40 47"
src="https://github.com/user-attachments/assets/b315424f-48d1-4d29-9e97-1fcf4d8c47f2"
/>
<img width="458" height="505" alt="Screenshot 2026-06-02 at 23 40 18"
src="https://github.com/user-attachments/assets/2a4e0ec2-8e0a-4220-84ec-177c788aa580"
/>
## Summary
Partners-app changes spanning the Partner object, its data scripts,
table views, and sidebar navigation.
### Remove the "Project Budget Typical" field
Dropped the `projectBudgetTypical` currency field from the Partner
object and every reference to it:
- `get-partner-by-slug` and `list-available-partners` logic-function
selections
- the seed script (type, write mapping, and per-partner data)
- the `import-from-tft` mapping (also dropping the now-unused
`partnerBudgetAverage` TFT source selection)
`projectBudgetMin` is intentionally kept.
### Rework partner views
- **Partners** (all-partners) view: replaced the **Deployment
Expertise** column with **Categories** (the `partnerScope` field).
- **Validated partners** view: added a **Languages Spoken** column.
- Set view `position`s so the in-object view switcher orders **Validated
→ Applications → Partners**.
### Navigation order
Reordered the "Partners" folder navigation items so the sidebar reads
**Validated partners → Partner applications → Partners** (Partner
content stays last).
### Also included
The previously-pushed fix that excludes partners with an empty slug from
the available-partners list.
## Notes
- No deploy/sync performed. The view-column and navigation-ordering
changes take effect once the app manifest is synced (`yarn twenty dev
--once` locally).
- The `deploymentExpertise` field itself is unchanged — only its column
was removed from the all-partners view.
Fixes https://github.com/twentyhq/twenty/issues/20840
## Summary
- Fix the AI tools factory (`view-filter-tools.factory.ts`) description
that incorrectly described SELECT filter values as plain strings,
causing agents to write values like `"CLOSED_LOST"` instead of the
required array format `["CLOSED_LOST"]`
This led to production crashes when later trying to update/delete select
options on fields that had view filters created by AI agents with
invalid format.
## Test plan
- No behavior change in existing code — only the tool description is
updated to guide AI agents correctly
- Manual: confirm AI agents now create SELECT filters with array values
## Why
`LambdaDriver` (1.4k lines) and `LocalDriver` (709 lines) each mixed
many unrelated concerns in a single class.
## What changed
**This PR only moves code** — every line is byte-for-byte the same as
`main` (same logic, same comments, same constants, same lock keys/TTLs,
same error mapping). The drivers are now thin orchestrators wiring
co-located sub-services + pure utils (each with unit tests).
| File | Before | After |
|---|---|---|
| `lambda.driver.ts` | 1425 | 252 |
| `local.driver.ts` | 709 | 231 |
| Largest sub-service | — | 371 (`lambda-executor-manager`) |
| Unit tests on these paths | 0 | 15 |
Lambda split: `lambda-aws-client` / `lambda-tool-functions` /
`lambda-layer-manager` / `lambda-executor-manager` + `constants` +
`types` + 5 pure utils.
Local split: `local-layer-manager` / `local-child-process-runner` /
`local-prebuilt-bundle` + `constants` + `types` + 3 pure utils.
No behavior change, no signature change, no public-API change.
Fixes https://github.com/twentyhq/twenty/issues/21000
Front-component event handlers read standard event fields
(event.clientX, event.offsetX, …), but these were always undefined. On
the remote side, serialized event data was passed only as the
CustomEvent's detail — and CustomEvent ignores every constructor option
except detail, so the values lived at event.detail.clientX and never on
the event object itself.
- Added `applySerializedEventProperties`, to copy a curated allowlist of
event-level keys onto the event. Element/target state (value, checked,
files, scroll, media props) stays in
`applySerializedEventTargetProperties`, applied to this (the dispatch
element = event.target).
- Added x/y to `SerializedEventData` and to host-side serialization in
`createHtmlHostWrapper`.
- Added an `svg-pointer `story + `createHtmlTagPointerStory`
Note: Also pinned @types/react to v18 so the renderer stops dragging in
React 19 types and breaking typecheck.
## Why
`WorkspaceMigrationBuildOrchestratorService.buildWorkspaceMigration` was
1117 lines of 30 near-identical copy-pasted blocks (one per metadata
entity) plus a 165-line ordered `actions: [...]` literal. Adding a new
entity meant copying ~30 lines and hoping the boilerplate stayed in
sync; the actual per-entity execution order was buried in the noise.
## Gain
- Orchestrator: **1117 → 388 lines**
- Net diff: **-558 lines**
- Adding a new entity is now a single line in the registry
- Execution order is visible at a glance
Combined two separate `TypeOrmModule.forFeature()` calls into one. Both
registered entities on the default data source, so no behavioral change.
Repositories for WorkspaceEntity and MessageChannelEntity remain
injectable as before.
Cleaner imports, one less redundant call.
## What
Lets a many-to-one relation be selected as the **leaf** of an advanced
(nested) filter. Previously the nested-field submenu excluded relations,
so you could filter `Opportunities WHERE company.Name contains X` but
not `Opportunities WHERE company.accountOwner = me`.
## How it works
Selecting a relation leaf filters by its **foreign key** —
`company.accountOwnerId = X` — a single hop the backend already resolves
on the joined table (`{ company: { accountOwnerId: { in: [...] } } }`).
It is **not** a multi-hop traversal: filtering on a *scalar field of*
the related record (e.g. `company.accountOwner.name`) stays excluded,
since that needs a second join the backend caps at one hop.
Two changes:
- **`AdvancedFilterRelationTargetFieldSelectMenu`** — stop excluding
many-to-one relations from the nested-field submenu.
- **`ObjectFilterDropdownRecordSelect`** — resolve the record picker's
object from the *leaf* relation's target (e.g. WorkspaceMember,
including the "Me" pin) rather than the source relation's object. The
source-field fallback applies only when there is no leaf.
## Testing
- Added `turnRecordFilterIntoRecordGqlOperationFilter` unit cases
asserting a relation leaf (and `= me`) compiles to the FK form — 59/59.
- typecheck + lint green (twenty-front, twenty-shared).
Seeding an onboarding view that uses this filter will follow in a
separate PR.
## Issue
- From Settings -> Admin Panel -> Workspace -> Members, impersonating
the currently logged-in user still issued an impersonation login token.
Token exchange produced invalid impersonation JWTs
(`impersonatorUserWorkspaceId === impersonatedUserWorkspaceId`). JWT
validation then failed with `User cannot impersonate themselves`,
leaving the app in an endless loading state until cookies were cleared.
- Closes#21086
## Approach
I was first thinking of to only hide the impersonate button for the
logged-in user in the admin, since they can not click what isn’t shown
(as I thought it was just a frontend issue).
But that was not enough:
- The `impersonate` mutation can still be called directly (GraphQL
client, scripts, devtools).
- Before this fix, the mutation could succeed and only fail later at JWT
validation, which led to invalid tokens and a broken session.
So the PR does both:
- Frontend: hide/disable self-impersonation in the UI and avoid
reloading on failed token exchange (UX).
- Backend: reject self-impersonation in `ImpersonationService` and at
token exchange (enforcement, fail fast before bad tokens).
Hiding the button is the right product behavior; the backend change is
what makes the rule real and safe.
## How to test
Manual:
- Log in as a user with admin impersonation.
- Go to Settings -> Admin panel -> Workspace -> open your workspace ->
members.
- Confirm your row has no Impersonate button; other members still do.
- Open Admin Panel -> User for yourself -> confirm no impersonate
button.
- Open Settings -> Members -> your own member profile -> confirm no
Impersonate action.
- Impersonate another member -> should work as before
Automated: `npx jest impersonation.service.spec`
### Before:
<img width="830" height="413" alt="Screenshot 2026-06-02 122224"
src="https://github.com/user-attachments/assets/46f38a74-8bd6-4ffa-b749-500ce18314f1"
/>
### After:
<img width="795" height="369" alt="Screenshot 2026-06-02 122333"
src="https://github.com/user-attachments/assets/62ece4a8-d38b-4f91-817f-792ff49b146b"
/>
---------
Signed-off-by: Parship Chowdhury <[email protected]>
Co-authored-by: Etienne <[email protected]>
Fixes https://github.com/twentyhq/twenty/issues/19899
Front component iframes were previously forced to `sandbox=""`, which
fully locks them down: no scripts, no forms, no popups. That broke any
legitimate embedded content (maps, widgets, embeds) developers tried to
render.
But we can't just trust the app-provided sandbox value either: tokens
like allow-same-origin or allow-top-navigation would let a malicious
embed escape the sandbox and hijack the host Twenty tab.
- Add `sanitizeIframeSandbox`, which keeps the iframe useful while
enforcing security:
applies a safe default (allow-scripts allow-forms allow-popups) when no
sandbox is set
always forces allow-scripts so embeds work
- strips dangerous tokens (`allow-same-origin`, all
`allow-top-navigation`*, `allow-popups-to-escape-sandbox`),
case-insensitively
- Wire it into `createHtmlHostWrapper` so every `iframe` rendered by a
front component is sanitized.
- Add unit tests for the sanitizer and Storybook interaction tests
asserting dangerous sandboxes are stripped.
## Summary
Iterative refinements to the partner design-doc doctrine after running
it on a second lead (TADA) and reviewing output side by side. Touches
only the `twenty-partner-design-doc` skill files (doctrine + Claude Code
wrapper); no runtime / app code.
**What changed**
- **Flag system:** emoji + short text label pairs only (`🔮 inf.`, **❓
open**, **⚠️ heavy**, **🛑 blocker**). Replaces the prior text-tag-only
system; scannable, unambiguous.
- **Section structure:** split into **Required** (always present) and
**Conditional** (Views, Automations, Integrations, Reporting). Include
conditional sections only when the client grounded them in the source.
Number sequentially, no gaps.
- **No filler placeholders:** banned `X was not named` / `left out on
purpose` lists in body sections. Unknowns belong in Open questions, not
as their own section or bullet.
- **Functional cross-refs:** every `§N` reference is now a markdown
anchor link `[§N](#n-section-slug)`, so a partner skimming the doc can
navigate. Bare `§N` is banned.
- **Bullets and tables over paragraphs**, with **Open questions** kept
as a numbered list (so the partner can read items 1, 2, 3 with the
client).
- **Views & navigation** rendered as a tight `Surface | Shows |
Audience` table. No view-type column — table / kanban / page layout is
the partner's call, not a scoping decision.
- **Data-model table** gains a `Source` column (`client` / `inf.`) for
at-a-glance fact-vs-inference visibility.
- **Business decisions over technical mechanics:** cut SDK / runtime
internals that don't move the quote (Docker version, OAuth flavour,
auto-system relations, env-var names, CI/CD workflow detail).
- **Common-mistakes table** updated with rows for the new rules.
- **SKILL.md self-check** expanded so the wrapper enforces all of the
above before saving.
## Test plan
- [ ] Re-read doctrine end-to-end for internal consistency
- [ ] Verify the four canonical emoji + text pairs appear and no stray
emoji flags remain
- [ ] Confirm Required vs Conditional structure is internally consistent
(no section listed in both)
- [ ] Confirm functional-cross-ref rule appears in both Rules and
Formatting and is reflected in the SKILL.md self-check
- [ ] Confirm Views & navigation entry mandates the three-column table
and bans a Type column
- [ ] Confirm Common-mistakes table covers each new rule
## What & why
The left-sidebar **Favorites** section was hidden whenever the user had
no favorites, so it was effectively undiscoverable — and personal
favorites could only be created via the record-level "Add to favorites"
action or drag-drop.
This PR:
- **Always shows the Favorites section**, with an empty-state **"Add a
favorite"** call-to-action.
- Adds a **"+" on the Favorites header that opens the same "New menu
item" side panel** the Workspace section already uses, so users can add
personal **Objects, Views, Records, Links and Folders** directly from
the sidebar.
## How
Favorites and workspace navigation are the same `NavigationMenuItem`
entity (a personal favorite simply has `userWorkspaceId` set). Rather
than build a separate favorites-only flow, the shared add/edit
side-panel subsystem is made **section-aware**
(`NavigationMenuItemSection = 'workspace' | 'favorite'`):
- a new `navigationMenuItemEditSectionState` atom records which section
the panel is operating on;
- a new `useNavigationMenuItemEditController` forks persistence — the
**workspace** section stages changes in the draft (saved on
layout-customization exit), while the **favorite** section
creates/updates/deletes personal items **immediately** with
`userWorkspaceId = current member`. This mirrors the existing
`useHandleNavigationMenuItemDragAndDrop` fork.
The existing add/edit hooks, pickers and title editors were rerouted
through the controller and a section-aware items hook, so they work for
both sections with no behavior change to the workspace flow.
**Backend: no changes** — `canUserCreateNavigationMenuItem` already
authorizes personal navigation menu items of every type for any
authenticated user.
## Decisions & tradeoffs
- **Folder button → unified "+":** the folder-only header button is
replaced by the single "+" (Folder is one of the panel's options),
matching the Workspace section. This removed the inline folder-create
code path.
- **Click-to-add only in v1:** dragging items from the panel directly
into Favorites is deferred — those drag handles are disabled in the
favorite section (the drag path is hardwired to workspace layout mode),
with a defense-in-depth no-op in the drop handler.
- **Persist-on-commit:** favorite title/URL edits hit the network once
on blur/enter, never per keystroke.
- **Personal color edits** change only the favorite's own color, never
the shared object metadata (that remains a workspace-customization
behavior).
- The change touches ~37 files because it generalizes the shared
subsystem rather than duplicating it; net diff is slightly negative
(+605 / −636).
## Testing
- `npx nx typecheck twenty-front` — passes
- `npx nx lint twenty-front` (oxlint + oxfmt) — passes
- `navigation-menu-item` unit tests — pass (incl. an updated
`computeInsertIndexAndPosition` test covering personal items)
- Manual end-to-end walkthrough still recommended before merge.
## What
Replaces `SubMenuTopBarContainer` with a settings-specific
`SettingsPageLayout` that puts the whole page chrome — breadcrumb,
centered title, actions, an optional secondary bar (tabs or wizard
step), and the 760px body — inside **one rounded card**, with
`SidePanelForDesktop` as a sibling. Title, tabs and body content share
one centered vertical axis at every card width.
Supersedes #21122. One PR, no feature flag.
## New components (`@/settings/components/layout/`)
- **SettingsPageLayout** — owns the rounded card + side-panel sibling,
`useCommandMenuHotKeys`, mobile command menu
- **SettingsPageHeader** — breadcrumb · centered title · actions in a
symmetric `1fr auto 1fr` grid (symmetric padding throughout)
- **SettingsSecondaryBar** — the secondary row, bracketed by top +
bottom borders
- **SettingsTabBar** — centered tabs reusing `activeTabIdComponentState`
+ `TabListFromUrlOptionalEffect` for URL-hash sync (does not touch the
shared `TabList`)
- **SettingsWizardStepBar** — back arrow · "N. Label" · optional
trailing slot
## Migrations
- Bulk rename across ~80 call sites (`SubMenuTopBarContainer` →
`SettingsPageLayout`); old component deleted.
- 5 tab pages (AI, APIs & Webhooks, Applications, Members, Role) + the
Data Model object-detail page render their tabs in `secondaryBar`
(object-detail keeps "See records" / "New Field" in the header actions).
- The 2 role object-level steps render the wizard step bar with working
back navigation.
- Accounts consolidated into **General / Emails / Calendars** tabs;
standalone `SettingsAccountsEmails` / `SettingsAccountsCalendars` pages
+ routes + stories removed. `SettingsPath.AccountsEmails` /
`AccountsCalendars` now resolve to `accounts#emails` /
`accounts#calendars`, so existing `getSettingsPath()` links deep-link to
the right tab via the existing hash sync — no call-site changes.
## Verification
- `nx typecheck twenty-front` and `nx lint twenty-front` both clean.
- Browser (logged-in workspace): title / tab / body / card centers align
on a single axis at multiple widths — width-invariant, so alignment
holds when the AI side panel (a sibling) shrinks the card. Rounded card
with even gaps on all four sides; tab row bracketed by two 1px lines;
no-tab pages render header → body with no lines; wizard back navigation
works; `…/accounts#emails` opens the Emails tab.
The shared `PageHeader` and `TabList` are untouched. The settings side
panel itself isn't wired to open yet — that's a follow-up PR.
Bumps the `twenty-partners` SDK app version 0.3.2 → 0.3.3 so `main`
tracks what's deployed to prod.
This is the deploy version for the partner-app changes that just landed:
marketplace `partnerScope` exposure (#21126), the
`submit-partner-application` endpoint + new Partner categories +
migration (#21040), the marketplace card rebind (#21127), and the signup
wizard (#21039).
No code changes — version bump only.
Fixes#20999
## Summary
Fixes a UX issue where clicking **Move left** or **Move right** in the
column header
dropdown immediately closed the menu, forcing users to reopen it for
every single move.
## Problem
`handleColumnMoveLeft` and `handleColumnMoveRight` both called
`closeDropdownAndToggleScroll()` unconditionally at the top of their
handlers — before
even checking `canMoveLeft` / `canMoveRight`. This immediately set the
Jotai atom
`isDropdownOpenComponentState` to `false`, unmounting the dropdown.
Since move actions are **repeatable** — a user might want to shift a
column several
positions — they were forced into a frustrating loop: click header →
click move → click
header → click move → repeat for every step.
## Fix
Removed the two `closeDropdownAndToggleScroll()` calls from the move
handlers in
`RecordTableColumnHeadDropdownMenu.tsx`.
```diff
const handleColumnMoveLeft = () => {
- closeDropdownAndToggleScroll();
-
if (!canMoveLeft) return;
moveTableColumn('left', recordField.fieldMetadataItemId);
};
const handleColumnMoveRight = () => {
- closeDropdownAndToggleScroll();
-
if (!canMoveRight) return;
moveTableColumn('right', recordField.fieldMetadataItemId);
};
```
All other handlers — **Filter, Sort, Hide** — are untouched and still
close the dropdown
correctly, since those are one-shot or navigation actions.
## Changes
| File | Change |
|---|---|
| `RecordTableColumnHeadDropdownMenu.tsx` | Remove 2
`closeDropdownAndToggleScroll()` calls from move handlers |
| `RecordTable.stories.tsx` | Add `HeaderMenuStaysOpenAfterMoveRight`
regression story |
## Testing
**Storybook interaction test** — `HeaderMenuStaysOpenAfterMoveRight`:
clicks "Move right" then asserts the menu is still visible.
**Manual checklist:**
- [x] Move right → menu stays open
- [x] Move right again → column moves again, menu still open
- [x] Move left → menu stays open
- [x] Move rightmost column → "Move right" disappears, menu stays open
showing "Move left"
- [x] Filter → menu closes *(unchanged)*
- [x] Sort → menu closes *(unchanged)*
- [x] Hide → menu closes *(unchanged)*
- [x] Click outside → menu closes *(unchanged)*
- [x] Escape → menu closes *(unchanged)*
- [x] TypeScript: zero new errors (`tsc --noEmit`)
---------
Co-authored-by: Félix Malfait <[email protected]>
Co-authored-by: Charles Bochet <[email protected]>
Fixes https://github.com/twentyhq/twenty/issues/21094
Conditional availability variables (`objectMetadataItem`,
`numberOfSelectedRecords`, `objectPermissions`, operators like
`everyEquals`/`none`, etc.) are compile-time-only constructs used in
`conditionalAvailabilityExpression`. They were previously exported from
`twenty-sdk/front-component`, which let developers mistakenly import
them into runtime component code where they have no value.
- Move conditional availability variables from
`twenty-sdk/front-component` to `twenty-sdk/define`.
- Add a build-time manifest validation
(validate-conditional-availability-usage) that fails the build if these
variables are imported/used outside of
`conditionalAvailabilityExpression`.
- Update the github-connector example app to register commands via
dedicated *.command-menu-item.ts files instead of inline command config
in front components.
- Update docs (all locales) and test mocks to reflect the new import
paths.
## Summary
### Why
1. Sending the code to the lambda (~1Mb usually) is heavy on network and
results to a constant traffic of ~30Mb/s on AWS which results into TB of
network data every month
2. eval(1MB of code) is not that fast, it's heavy on memory and CPU on
lambda side
### High level
Adds two execution modes for logic functions, gated behind the new
`IS_LOGIC_FUNCTION_PREBUILT_MODE_ENABLED` workspace feature flag (off
everywhere by default):
- **LIVE** (current behavior, preserved bit-for-bit): the compiled
bundle is read from object storage and shipped in every Lambda invoke
payload. Used for fast iteration in the workflow editor / Settings test
runs.
- **PREBUILT** (new): the bundle is installed onto the per-function
Lambda alongside the unified executor, and invocations carry only `{
params, env, handlerName }` — saving JSON payload egress and warm-start
`import()` cost on every call.
### Key design choices
- **Unified Lambda handler** (`constants/executor/index.mjs`) dispatches
at runtime: `event.code` present ? LIVE (write to `/tmp`, dynamic
import) : `import('./prebuilt-logic-function.mjs')`. Both code paths
always coexist on the deployment package, so the same Lambda can serve
either mode without redeploying.
- **Install runs inside the `validateBuildAndRun` migration pipeline**,
not at execute time. `Create/UpdateLogicFunctionActionHandlerService`
calls `driver.installPrebuiltBundle` when `executionMode` flips
LIVE?PREBUILT or `checksum` changes while PREBUILT, gated on
`isBuildUpToDate=true` and a fresh checksum.
- **Strict execute, no reconciliation**:
`LogicFunctionExecutorService.execute` resolves `effectiveExecutionMode`
(caller override > feature flag > entity column). For PREBUILT it asks
the driver `getInstalledBundleChecksum` (Lambda `twenty:bundle-checksum`
tag for AWS, sidecar file locally) and throws
`LOGIC_FUNCTION_PREBUILT_BUNDLE_NOT_INSTALLED` on mismatch.
- **Feature flag gates every side effect**: with the flag off the
executor forces LIVE, the action-handler install hooks bail before AWS,
and workflow activation does not flip the mode. Rollback is just turning
the flag off.
### Lifecycle
- New workflow CODE step ? `LIVE`, no install.
- Workflow activated ? build + activation flips `executionMode=PREBUILT`
? action-handler installs the bundle + sets the Lambda tag.
- Draft from active version ? duplicated logic function reset to `LIVE`.
- App install ? manifest converter sets `PREBUILT`, create-action
handler installs.
- Test runs (`executeOneFromSource`, workflow editor) pass
`executionMode=LIVE` explicitly.
### Observability
`[lambda-timing]` log lines now include `effectiveExecutionMode` and
`payloadBytes`; the action handler logs `install_duration_ms` for each
install.
## Test plan
- [x] `npx nx typecheck twenty-server` ? passes
- [x] `npx oxlint --type-aware` on all changed files ? 0 warnings, 0
errors
- [x] `npx nx test twenty-server` ? 588 suites / 5009 tests pass (no
regressions vs main)
- [x] New unit suite `flat-logic-function-validator.service.spec.ts` ?
9/9
- [x] Existing
`workflow-version-step-operations.workspace-service.spec.ts` ? 8/8
(verified the new token-based DI avoids a circular-import regression)
- [x] Snapshot for
`ALL_UNIVERSAL_FLAT_ENTITY_PROPERTIES_TO_COMPARE_AND_STRINGIFY` updated
to include `executionMode`
- [x] Integration suite `logic-function-execution.integration-spec.ts`
extended to assert `executionMode=LIVE` on newly-created functions and
continues to exercise the LIVE happy path
- [ ] Manual staging rollout: flip
`IS_LOGIC_FUNCTION_PREBUILT_MODE_ENABLED` per workspace, observe
`[lambda-timing]` `payloadBytes` drop + `install_duration_ms`, then ramp
in prod.
## Problem
The AI `find_<object>` tool builds its input schema with
`generateFindToolInputSchema`, which **spreads field filters at the args
root** (alongside `limit`/`offset`/`orderBy`/`and`/`or`/`not`).
`tool-executor.service.ts` then maps the raw model args to a filter
with:
```ts
const { limit, offset, orderBy, ...filter } = args;
```
The zod schema is only used to generate the JSON schema *shown* to the
model (`z.toJSONSchema(...)`) — it is **never used to validate the args
coming back**. So when the model emits a bare operator where a field
name belongs, e.g. `{ ilike: "Foreman" }`, it passes straight to the
query runner, which throws and burns a retry mid-turn:
```
ERROR [FindRecordsService] Failed to find records: Object person doesn't have any "ilike" field.
ERROR [FindRecordsService] Failed to find records: Object person doesn't have any "eq" field.
```
Two contributing faults:
1. **The tool description actively misleads the model** — it says ``use
filter: { id: { eq: "record-id" } }``, a `filter` wrapper the
root-spread schema doesn't have, inviting the malformed shape.
2. **No server-side validation** — invalid root keys reach the query
runner instead of being rejected against the advertised contract.
## Fix
1. **`FindRecordsService` prunes invalid filter keys before querying.**
Using the same filter shape the tool schema advertises
(`generateRecordFilterSchema(...).filterShape`), it drops any key that
is neither a real field nor a logical operator (`and`/`or`/`not`),
recursing through `and`/`or`/`not`. A model that sends `{ ilike:
"Foreman" }` now gets a valid (empty) filter rather than an exception.
Extracted as a pure, unit-tested util `pruneFilterToAllowedKeys`.
2. **Corrected the `find_<object>` tool description** to describe the
real top-level-field shape and explicitly warn against a `filter`
wrapper and bare root operators.
## Test
`__tests__/prune-filter-to-allowed-keys.util.spec.ts` covers: valid
filters untouched, bare root operators dropped, valid siblings
preserved, `and`/`or`/`not` recursion, and non-object input.
## Notes
- Defensive for all `FindRecordsService` callers; `find_one` (`{ id: {
eq } }`) and workflow find-records pass valid filters and are
unaffected.
- Companion to #21106 (RICH_TEXT composite filters). Both surfaced from
the same `"Tom Foreman's notes"` AI-chat repro; this PR addresses the
root-level-operator half.
---------
Co-authored-by: Rich Roberts <[email protected]>
## Description
This PR resolves a usability issue where scalar field values (emails,
phone numbers, dates, IDs, text, etc.) rendered in display-mode or
read-only mode in the record detail side panel could not be highlighted,
selected, or copied natively with the cursor.
## Root Cause
Both `RecordInlineCellContainer` and
`RecordInlineCellHoveredPortalContent` wrapper elements had
`user-select: none;` hardcoded in their styled-component definitions.
This styling propagated down to all nested display widgets, locking
their content and preventing native text selection.
## Changes
- Updated `StyledInlineCellBaseContainer` in
`RecordInlineCellContainer.tsx` to use `user-select: text;` instead of
`none;`.
- Updated `StyledInlineCellBaseContainer` in
`RecordInlineCellHoveredPortalContent.tsx` to use `user-select: text;`
instead of `none;`.
These changes restore natural browser text selection capabilities for
record detail widgets without altering interactive edit-mode behaviors.
## Verification
- Verified styling changes.
- Tested locally to ensure that text highlighting and copy-pasting
function correctly when dragging over read-only fields.
Closes#21056
---------
Co-authored-by: Charles Bochet <[email protected]>
## Summary
Adds a public `POST /partner-applications` HTTP logic function on the
twenty-partners SDK app that receives applications from the website
wizard and idempotently upserts the Partner / Person / Company graph in
the partners workspace. Also introduces the validated **Category**
taxonomy on `partnerScope` (additive, prod-safe) plus the legacy→new
migration tooling.
Companion PR (website side): #21039
### Logic function
- `defineLogicFunction({ httpRouteTriggerSettings: { path:
'/partner-applications', httpMethod: 'POST', isAuthRequired: false,
forwardedRequestHeaders: ['x-application-secret'] } })`.
- Authenticates via shared-secret header (`X-Application-Secret` ↔
`PARTNER_APPLICATION_SECRET` workspace variable). Twenty's
`isAuthRequired: true` only accepts user-session JWTs, so the handler
enforces auth itself.
- Idempotent upsert keyed on `Person.emails.primaryEmail`:
- missing email → create Company → Person → Partner
- existing Person, no Partner → create Company + Partner, link
- existing Person + Partner → update Partner fields; preserve
staff-owned columns (`validationStage`, `reviewed`, `ranking`,
`partnerTier`, `lastMatchAt`) by omitting them from the update
- Create-time defaults preserved on resubmit: `slug =
slugify(companyName)` ("YC Agency" → "yc-agency"), `reviewed = false`,
`partnerTier = 'NEW'`.
- Currency conversion to `{ amountMicros, currencyCode: 'USD' }` for
`hourlyRate` + `projectBudgetMin`.
### Categories (`partnerScope`) — additive, prod-safe
- Adds 5 validated category options — `ADVISORY`, `SOLUTIONING`,
`DEVELOPMENT`, `HOSTING`, `SUPPORT` — to the `partnerScope` MULTI_SELECT
**without removing** the legacy options (there is production data on
them). Field relabeled **"Categories"**. The website form only emits the
new values.
- **Migration tooling** (run deliberately, *not* in CI):
`scripts/migrate-partner-scope.ts` remaps existing records legacy→new —
dry-run by default, `MIGRATE_APPLY=1` to write, two-pass
(collect-then-apply, no mutate-while-paginating).
`scripts/partner-scope-map.ts` is the single mapping source;
`import-from-tft.ts` now routes imported scope through it so the TFT
import never re-introduces retired values. Removing the legacy options
is deferred until after the migration has run + been verified.
### applicationNotes
- New `applicationNotes` TEXT field holds the wizard's single free-text
"anything else" note (the handler passes it through directly).
`deploymentExpertise` was dropped from the handler
input/validation/builders (the column is retained for now, pending the
same migration cleanup).
### Application variable
- Declares `PARTNER_APPLICATION_SECRET` with `isSecret: true` so each
workspace sets the value via Settings → Apps → Twenty Partners →
Variables. Twenty encrypts at rest and merges the decrypted value into
the handler's `process.env` at execution time (workspace value wins over
container env).
### Code quality (from review)
- One shared `slugify` (`scripts/slugify.ts`, the import's algorithm)
used by both the handler and the import, so the `slug` identity key
can't diverge across paths.
- Unit-test tier: `vitest.unit.config.ts` (no `globalSetup`) + `yarn
test:unit`, so the pure `mapLegacyScope` test runs without a live server
(the integration suite stays server-backed).
## Demo
📹 _Screen recording of the wizard end-to-end (open → walk steps → submit
→ Partner record lands):_
https://github.com/user-attachments/assets/7458dd86-e3ff-47b5-9878-0eb134ff38e3
### Tests
- Integration tests against a local Twenty workspace:
missing-/wrong-secret auth rejections, create flow (asserts slug +
`reviewed: false` + `partnerTier: 'NEW'`), update-on-resubmit +
staff-column preservation, new category values stored,
`applicationNotes` stored, bad-input shape.
- Pure `mapLegacyScope` unit test via `yarn test:unit` (no server).
## Test plan
- [ ] Install / upgrade the app on the target workspace; set
`PARTNER_APPLICATION_SECRET` in Settings → Apps → Twenty Partners →
Variables
- [ ] `curl -i -X POST <workspace-url>/s/partner-applications -H
'X-Application-Secret: <secret>' -H 'Content-Type: application/json' -d
'{"firstName":"Test","lastName":"User","email":"[email protected]","companyName":"YC
Agency","partnerScope":["ADVISORY"],"applicationNotes":"hi"}'` →
`HTTP/1.1 201` + `{"ok":true,"created":true,"partnerId":"..."}`
- [ ] Partner record shows `name: "YC Agency"`, `slug: "yc-agency"`,
`validationStage: APPLICATION`, `reviewed: false`, `partnerTier: 'NEW'`,
`partnerScope: ["ADVISORY"]`, `applicationNotes: "hi"`
- [ ] Re-curl same email with `city: "Paris"` → `created: false`,
`Partner.city` updated, staff-owned columns untouched
- [ ] Wrong / missing secret → `200` +
`{"ok":false,"reason":"unauthorized"}`
- [ ] `yarn test:unit` green (no server); `yarn migrate:partner-scope`
dry-run lists any legacy→new remaps without writing
Charles Bochet