fix(data-fabric): MULTILINE_MAX list reads are a preview, not always a marker

skills/uipath-coded-apps/references/sdk/data-fabric.md

@@ -18,7 +18,7 @@ Data Fabric does NOT behave like a typical RDBMS. These server behaviors are inv
 6. **No `IsNull` filter operator.** The server can't ask "where field is null" (`QueryFilterOperator` has no such member). Filter client-side after fetching, or design with explicit non-null sentinels.
 7. **Aggregates require server-side `aggregates` + `groupBy`.** Don't fetch raw rows and `.length` / `.reduce` client-side — every list call returns one page (see [pagination.md](pagination.md)), so `result.items.length` after `queryRecordsById({ filter })` returns at most one page's worth, no matter how many rows match. Use `totalCount` for cardinality, `aggregates: [{ function: 'COUNT', field: 'Id' }]` (with `groupBy` for per-bucket counts) for chart data.
 8. **File-type fields (`fieldDisplayType === 'File'`) aren't strings.** The record carries only metadata (`{ id, name, size, contentType }`); stringifying gives `"[object Object]"`. To display, call `entities.downloadAttachment(entityId, recordId, fieldName)` → `Blob` → `URL.createObjectURL` for an `<img src>`. **Neither `contentType` nor filename extension is reliable for detecting kind** — DF often returns `application/octet-stream`, and the stored `name` is frequently a bare UUID with no extension. To decide inline render vs download link, either (a) sniff the blob's magic bytes after download (PNG starts `89 50 4E 47`, JPEG `FF D8 FF`, GIF `47 49 46 38`, PDF `25 50 44 46`), or (b) optimistically attempt `<img src={objectUrl}>` and swap to a download link in `onError`. Writes go through `uploadAttachment(...)`, not `insertRecordById` / `updateRecordById`.
-9. **`MULTILINE_MAX` fields return a size marker on list/query reads.** `getAllRecords` / `queryRecordsById` return a string starting `HasValue=true Length=N` (live form: `"HasValue=true Length=20000 — call Get Entity Record By Id activity to retrieve content"`), never the content — only `getRecordById` returns the full value (SDK 1.5.2+, v2 read endpoint). Never render or persist the marker as data, and never echo it back through `updateRecordById` / `updateRecordsById` — the server accepts it as a normal value and silently destroys the real content; omit the key instead. The type accepts no filters or `sortOptions` (server 400: *"Field '<name>' is of type MULTILINE_MAX and cannot be used in filters."*). `lengthLimit` is a UTF-16 **byte** budget (max 131072 ≈ 65,536 chars).
+9. **`MULTILINE_MAX` fields return only a preview on list/query reads.** `getAllRecords` / `queryRecordsById` return either the value truncated to 10,000 characters with a `...[Truncated]` suffix or a size marker starting `HasValue=true Length=N` (sometimes with a trailing hint), depending on the tenant; encrypted fields always return `HasValue=true Encrypted=true`. Only `getRecordById` returns the guaranteed full value (SDK 1.5.2+, v2 read endpoint). Never render or persist a preview as data, and never echo it back through `updateRecordById` / `updateRecordsById` — the server accepts it as a normal value and silently destroys the real content; omit the key instead. The type accepts no filters or `sortOptions` (server 400: *"Field '<name>' is of type MULTILINE_MAX and cannot be used in filters."*). `lengthLimit` is a UTF-16 **byte** budget (max 131072 ≈ 65,536 chars).
 
 ## Three paths require choice-value translation — don't miss any
 

skills/uipath-platform/references/data-fabric/data-fabric.md

@@ -111,7 +111,7 @@ Do not change field types, create federated entities, or write federated records
 
 20. **`records import` supports Basic types only.** `CHOICE_SET_*`, `RELATIONSHIP`, `FILE`, `AUTO_NUMBER` columns are ignored on import — optional columns land as `null`; **`isRequired` columns without a `defaultValue` fail the whole row** (`ErrorFileLink` entry per row). Sequence: (1) `entities get` → list unsupported columns; (2) tell the user which columns will be skipped, and which rows will fail because a required unsupported column has no default; (3) offer `records insert --file <json>` (+ `files upload` for FILE) as the alternative; (4) invoke only after explicit confirmation. See [`bulk-import.md` → Complex Field Types Not Supported](bulk-import.md#complex-field-types-not-supported).
 
-21. **`MULTILINE_MAX` — marker reads, no filter/sort.** `records list` / `records query` return a size marker (`HasValue=true Length=N`) — full value only via `records get`. Never echo the marker back through `records update` — the server accepts it as a normal value and destroys the real content; omit the key instead. No filter/sort support (400). `lengthLimit` is a UTF-16 byte budget (max 131072 ≈ 65,536 chars). Full contract: [`entity-schema.md` → MULTILINE_MAX](entity-schema.md#multiline_max-fields) + [`records-query.md` → MULTILINE_MAX](records-query.md#multiline_max-fields--marker-vs-full-content).
+21. **`MULTILINE_MAX` — preview reads, no filter/sort.** `records list` / `records query` return a preview, either content truncated to 10,000 characters with a `...[Truncated]` suffix or a size marker (`HasValue=true Length=N`), depending on the tenant; full value only via `records get`. Never echo a preview back through `records update` — the server accepts it as a normal value and destroys the real content; omit the key instead. No filter/sort support (400). `lengthLimit` is a UTF-16 byte budget (max 131072 ≈ 65,536 chars). Full contract: [`entity-schema.md` → MULTILINE_MAX](entity-schema.md#multiline_max-fields) + [`records-query.md` → MULTILINE_MAX](records-query.md#multiline_max-fields--preview-vs-full-content).
 
 ---
 

skills/uipath-platform/references/data-fabric/entity-schema.md

@@ -5,7 +5,7 @@
 > **Preview-then-confirm gate (data-fabric.md Rule 14).** Before invoking `entities create` — or any `entities update` that adds, updates, or removes fields — render the full proposed schema (entity name, displayName, description, every field with normalized type and all extras) as a table or formatted JSON block and wait for explicit user approval. Don't run the CLI until the user confirms.
 
 ```bash
 uip df entities create "MyEntity" \
   --body '{
     "displayName": "My Entity",
     "description": "Optional description",
@@ -32,7 +32,7 @@
 |---|---|---|
 | `STRING` | NVARCHAR | Short text (≤4000 chars via `lengthLimit`) |
 | `MULTILINE_TEXT` | NVARCHAR(MAX) | Long text (≤10000 chars via `lengthLimit`) |
-| `MULTILINE_MAX` | NVARCHAR(MAX) | Very large text (`lengthLimit` = UTF-16 byte budget, 1–131072; default 128 KB ≈ 65,536 chars max). No filter/sort; list/query reads return a size marker — see [MULTILINE_MAX fields](#multiline_max-fields) |
+| `MULTILINE_MAX` | NVARCHAR(MAX) | Very large text (`lengthLimit` = UTF-16 byte budget, 1–131072; default 128 KB ≈ 65,536 chars max). No filter/sort; list/query reads return a preview, not the guaranteed full value — see [MULTILINE_MAX fields](#multiline_max-fields) |
 | `DECIMAL` | DECIMAL | All numbers — `decimalPrecision: 0` for whole; `2` for money |
 | `BOOLEAN` | BIT | true/false |
 | `DATE` | DATE | Date only |
@@ -80,11 +80,11 @@
 Very large text. Contract differs from `MULTILINE_TEXT`:
 
 1. **Not filterable, not sortable.** Any `queryFilters` or `sortOptions` entry naming a `MULTILINE_MAX` field → 400: *"Field '<name>' is of type MULTILINE_MAX and cannot be used in filters."* / *"Sort field '<name>' is of type MULTILINE_MAX and cannot be used for sorting."* Never offer the field in filter/sort predicates. See [filter contract](filter-platform-contract.md#operator-support-by-field-type).
-2. **List/query reads return a size marker, not content.** `records list` / `records query` return a string starting `HasValue=true Length=N` (live form: `"HasValue=true Length=20000 — call Get Entity Record By Id activity to retrieve content"`); only `records get <entity-id> <record-id>` returns the full value. Read + write-back rules in [records-query.md → MULTILINE_MAX fields](records-query.md#multiline_max-fields--marker-vs-full-content).
+2. **List/query reads return a preview, not the guaranteed full value.** Depending on the tenant, `records list` / `records query` return either the value truncated to 10,000 characters with a `...[Truncated]` suffix, or a size marker starting `HasValue=true Length=N` (sometimes with a trailing hint). Encrypted fields always read back `HasValue=true Encrypted=true`. Only `records get <entity-id> <record-id>` returns the full value. Read + write-back rules in [records-query.md → MULTILINE_MAX fields](records-query.md#multiline_max-fields--preview-vs-full-content).
 3. **On a 400 from `entities create` / `addFields` naming the type**, surface the error verbatim — do NOT retry or silently substitute `MULTILINE_TEXT` (Rule 18).
 
 ```bash
 uip df entities create "Documents" \
   --body '{"fields":[{"name":"Title","type":"STRING","isRequired":true},{"name":"Body","type":"MULTILINE_MAX"}]}' \
   --output json
 ```
@@ -161,7 +161,7 @@
 | `decimalPrecision` | `DECIMAL` — `0` whole, `2` money | 0–10 |
 
 ```bash
 uip df entities create "Orders" \
   --body '{
     "fields": [
       {"name": "ProductName", "type": "STRING",  "lengthLimit": 500, "isRequired": true},

skills/uipath-platform/references/data-fabric/records-query.md

@@ -16,16 +16,21 @@ Response wrapper: `{ Result, Code: "RecordList" | "RecordQuery", Data: { Items,
 - **`Data.NextCursor` is an object `{ "Value": "<base64-string>" }`, not a flat string.** Pass `Data.NextCursor.Value` to `--cursor` on the next call (unwrap one level). Passing the whole `NextCursor` object errors out.
 - Use `Data.HasNextPage` to check if more records exist. Stop when it's `false`.
 
-## MULTILINE_MAX Fields — Marker vs Full Content
+## MULTILINE_MAX Fields — Preview vs Full Content
 
-`records list` and `records query` do NOT return `MULTILINE_MAX` content. Each such field comes back as a size marker string starting `HasValue=true Length=N` — live form: `"HasValue=true Length=20000 — call Get Entity Record By Id activity to retrieve content"`. Only single-record read returns the full content:
+`records list` and `records query` do NOT reliably return `MULTILINE_MAX` content. Each such field comes back as a **preview**, and which form you get depends on the tenant:
+
+- the value truncated to 10,000 characters with a `...[Truncated]` suffix (a value at or under 10,000 characters can come back whole), or
+- a size marker starting `HasValue=true Length=N`, sometimes with a trailing hint such as `"HasValue=true Length=20000 — call Get Entity Record By Id activity to retrieve content"`.
+
+An encrypted `MULTILINE_MAX` field always reads back as `HasValue=true Encrypted=true` and never returns content. Do not branch on which form you got; treat any of them as a preview. Only the single-record read returns the full value:
 
 ```bash
 uip df records get <entity-id> <record-id> --output json
 ```
 
-1. **Never treat the marker as the value.** Don't display, compare, or persist `"HasValue=true Length=N"` as field content — fetch via `records get` first.
-2. **Never write the marker back.** A `records update` body built by echoing a record from `list` / `query` overwrites the real content with the literal marker string — verified: the server accepts it as a normal value, `Result: Success`, content silently destroyed. Omit `MULTILINE_MAX` keys from update bodies unless intentionally replacing the content.
+1. **Never treat a preview as the value.** Don't display, compare, or persist what `list` / `query` returned for the field; fetch via `records get` first. A truncated preview is the more dangerous form because it looks like ordinary content.
+2. **Never write a preview back.** A `records update` body built by echoing a record from `list` / `query` overwrites the real content with the preview — verified: the server accepts it as a normal value, `Result: Success`, content silently destroyed. With a truncated preview the stored value keeps its first 10,000 characters and loses the rest, which is far harder to notice than a marker string. Omit `MULTILINE_MAX` keys from update bodies unless intentionally replacing the content.
 3. **No filter, no sort.** `queryFilters` / `sortOptions` naming a `MULTILINE_MAX` field → 400: *"Field '<name>' is of type MULTILINE_MAX and cannot be used in filters."* / *"Sort field '<name>' is of type MULTILINE_MAX and cannot be used for sorting."* Surface verbatim (data-fabric.md Rule 18); don't retry with other operators. Full type contract: [entity-schema.md → MULTILINE_MAX fields](entity-schema.md#multiline_max-fields).
 
 ## Pagination

tests/tasks/uipath-platform/data-fabric/integration_multiline_max.yaml

@@ -1,10 +1,10 @@
 task_id: skill-datafabric-integration-multiline-max
 description: >
   Integration: MULTILINE_MAX lifecycle — create an entity with a MULTILINE_MAX
-  field, insert ~20k-char content, observe the size marker on records list,
-  retrieve full content via records get, and surface the server-side filter
-  rejection verbatim without retrying (Rules 18 + 21). Requires the
-  MultilineMax feature flag enabled on the test tenant.
+  field, insert ~20k-char content, observe that records list returns only a
+  preview of it, retrieve the full content via records get, and surface the
+  server-side filter rejection verbatim without retrying (Rules 18 + 21).
+  Requires the MultilineMax feature enabled on the test tenant.
 tags: [uipath-platform, data-fabric, integration, entities, records, mode:build, lifecycle:setup]
 
 run_limits:
@@ -72,25 +72,34 @@ success_criteria:
     pass_threshold: 1.0
 
   - type: llm_judge
-    description: "Agent explained the marker-vs-full contract and surfaced the filter rejection without silent workarounds"
+    description: "Agent explained the preview-vs-full contract and surfaced the filter rejection without silent workarounds"
     include_agent_output: true
     prompt: |
       The agent worked against a Data Fabric entity whose Body field is
       MULTILINE_MAX. Platform contract: `records list` / `records query`
-      return a size marker string starting "HasValue=true Length=N" (with a
-      trailing hint about Get Entity Record By Id) for such fields; only
-      `records get` returns the full content; server-side filters on the
+      return only a PREVIEW of such a field, never the guaranteed full value.
+      The preview takes one of two forms depending on the tenant, and BOTH are
+      correct platform behaviour:
+        - the content truncated to 10,000 characters with a "...[Truncated]"
+          suffix, or
+        - a size marker string starting "HasValue=true Length=N", sometimes
+          with a trailing hint about Get Entity Record By Id.
+      Only `records get` returns the full content. Server-side filters on the
       field are rejected with HTTP 400 ("Field 'Body' is of type
       MULTILINE_MAX and cannot be used in filters.").
 
       Read the agent's final response. Score on three points:
 
-      a) The agent showed that `records list` returned a size marker (not the
-         content) for Body, and showed the full ~20,000-char content (or a
+      a) The agent showed what `records list` returned for Body and it was a
+         preview in either of the two forms above (i.e. not the full ~20,000
+         characters), and showed the full ~20,000-char content (or a
          clearly-labelled excerpt of it) obtained via the single-record read.
-      b) The agent explained the difference correctly (list/query return a
-         marker; the single-record read returns full content) — not a wrong
-         theory like truncation, encoding, or a bug.
+      b) The agent explained the difference correctly: list/query give a
+         bounded preview and the single-record read gives the full content.
+         Describing the preview as truncated content, or as a size marker,
+         is correct as long as it matches what the tenant actually returned.
+         Score this point false only for a wrong theory, such as claiming
+         list returns the complete value, or blaming encoding or a bug.
       c) For the "contains lorem" filter request, the agent either surfaced
          the server's rejection verbatim (real error text/code) or stated
          up-front that MULTILINE_MAX is not filterable citing the platform