diff --git a/specs/epics/FIX-2516/DECISIONS.md b/specs/epics/FIX-2516/DECISIONS.md
new file mode 100644
index 0000000000..0fe5ceb4f8
--- /dev/null
+++ b/specs/epics/FIX-2516/DECISIONS.md
@@ -0,0 +1,177 @@
+# FIX-2516 · Decisions
+
+[Spec](SPEC.md) · **Decisions**
+
+The calls that sit above any single child: what was chosen, what lost, and what each locks in
+for the eight children and the closure. D1 is the sign-off. Q1 and Q2 are open, and they are the
+owner's, because each sets a stored format or a public contract. The engineering questions the
+brief raised are answered under *Decided, not asked*.
+
+## The tree
+
+```mermaid
+flowchart TD
+ E["FIX-2516"] --> Q1["Q1 · open · stored sessions keep the bytes inline in v1"]
+ Q1 -.->|"recommended against"| X1["references to a blob store, FIX-824 first"]
+ E --> Q2["Q2 · open · attachments ride beside the input"]
+ Q2 -.->|"recommended against"| X2["inside each flow's input schema"]
+ E --> D1["D1 · eight children and a closure, images then documents"]
+ D1 -.->|"rejected"| X3["documents folded into the framework and harness issues"]
+```
+
+
+## Q1 · Open · In v1, do stored sessions keep each attachment's bytes inside the message?
+
+**In plain terms.** A sent image must live somewhere so a reload can show it and later turns can
+see it. Either the message keeps the bytes, or it keeps a pointer to a file store we don't have.
+
+**The options.** *Inline:* the request carries each file as base64, and the stored user line
+keeps it in contracts' existing `FileContent.data`. *Reference:* build FIX-824's upload boundary
+and FIX-367's binary store first, and store a pointer. *Strip:* mark the bytes `ephemeral`.
+
+**The trade-off.** Inline ships now on a stored shape that exists, and it is what the AI SDK's
+chat UI does. Its price is size: a 10-image turn adds about 3 MB to the session and to every
+read of it. It also sets the ceiling. Vercel, where Kitchen Sink runs, refuses a request over
+4.5 MB, so a message carries at most 3 MB of files after downscaling: enough for 10 screenshots
+at 1,568 px, but a PDF caps at 3 MB, not the brief's 20 MB. Reference has neither cost and puts
+another epic first. Strip loses the reload leg.
+
+**My recommendation: inline, capped, measured first.** FIX-2517's first PR measures the kill
+line: 10 real screenshots through the downscale rule against 4.5 MB, and a session with three
+such turns reloaded in under two seconds on a local store.
+
+**What would change my mind.** That measurement failing. Then the kill line trips.
+
+**What being wrong costs.** Inline, wrongly: stored sessions hold bytes, and the later move to
+pointers reads both shapes (BP-030). Reference, wrongly: the epic waits an epic for a store that
+10 screenshots didn't need.
+
+
+
+It comes down to the reload: only inline shows the image without building a store first.
+
+
+## Q2 · Open · Do attachments ride beside the action's input, or inside it?
+
+**In plain terms.** Either every request gets one framework field for files, or each flow puts
+files in its own input and checks them itself.
+
+**The options.** *Beside:* `sendAction(action, input, { attachments })`. The engine checks the
+limits and writes the files as `file` parts on the user line the action's `userMessage` writes;
+a block reads them there. An action with no `userMessage` refuses files with a named error.
+`HarnessRunInput` gains the same list. *Inside:* each flow declares and checks its own field.
+
+**The trade-off.** Beside, every flow, worker door and harness turn takes files with no schema
+change, behind one server check. Its price is public surface: a client and React option, a
+request field, a block read, a harness input field. Inside adds none, and is FIX-824's problem
+again: every app re-implements the field, the check and the rendering.
+
+**My recommendation: beside.** Only there can't a hand-built request skip the limits.
+
+**What would change my mind.** Files that are not part of what the person said, such as a batch
+import. That is FIX-137's broader problem; it would sit beside this field.
+
+**What being wrong costs.** Beside, wrongly: a field we deprecate in a minor. Inside, wrongly:
+five apps carry their own copy, which FIX-2369 ER-3 forbids.
+
+
+
+It comes down to the server check: inside the input, every flow writes its own or skips it.
+
+
+## D1 · Eight children and a closure: the framework and the composer in parallel, then the harnesses, then the apps; documents after images
+
+| | |
+|---|---|
+| **Instead of** | Documents folded into FIX-2517 and FIX-2519 · one issue for both labs |
+| **Because** | Images reach every target the same way. Documents don't: Codex and Cursor take no documents, so each target needs its own answer, and that answer should not hold images back. Each app has its own composer and its own send path, so an app issue is the unit a reviewer can check in a browser |
+| **Locks in** | Nine rows. A collapse trigger on the labs: if both lab specs come out with no decision of their own, the spec set merges them |
+
+**What would change my mind:** the two lab specs being identical. Then they are one issue.
+
+
+
+It comes down to Codex and Cursor: with no document input, documents need their own answer.
+
+## Who owns what
+
+
+
+Each rule has one owner. A *uses* cell is a place where a child must not decide again.
+
+| # | Rule every child obeys | Owner | Checked at |
+|---|---|---|---|
+| ER-1 | One attachment shape: contracts' `FileContent` (`type: "file"`, `mediaType`, `data`, `filename`) on the user item and the same parts in the request. No second shape and no `attachment` item type | FIX-2517 | Spec set; cross-spec review |
+| ER-2 | One limits set, defined once in core, read by the composer's defaults and by the server. No app sets its own | FIX-2517 | Spec set; the closure's leg c |
+| ER-3 | Nothing over a limit is dropped silently. The composer refuses before send with a plain message; the server refuses with a named error before any block runs | FIX-2517 names the refusals · FIX-2518 uses them in the composer | Each child's goal check; leg c |
+| ER-4 | A harness or provider that can't take a type says so in the conversation; never a text-only fallback | FIX-2519 | FIX-2519's and FIX-2520's goal checks |
+| ER-5 | Attachments ride beside the input ([Q2](#q2)). The worker door's `{ message }` and each flow's input schema stay as they are | FIX-2517 | Spec set |
+| ER-6 | Composer pieces come from the `packages/ui` registry through `fsdev ui add`. No app writes its own paste, drop, preview or downscale (FIX-2369 ER-3, FIX-2152 ER-3 and ER-4) | FIX-2518 | Each app's spec; the closure's sweep |
+| ER-7 | A task's attachments are part of its brief, on the task row in orchestration, in [Q1](#q1)'s stored form. Every run of the task gets them. No task noun enters core (FIX-2369 ER-1) | FIX-2522 · FIX-2519 uses it to hand them to each run | Both specs; the closure's Shift Manager re-run |
+| ER-8 | Files written to disk for a harness live in a per-turn directory outside the checkout, and the adapter that wrote it removes it when the turn ends. Nothing reaches a commit | FIX-2519 | FIX-2519's goal check |
+| ER-9 | The kill line is measured in FIX-2517's first PR, before any app child starts. If it trips, the coordinator stops FIX-2521 to 2524 and amends this epic | FIX-2517 · the coordinator uses the result | FIX-2517's first PR |
+| ER-10 | Consumed issues keep their own parents: FIX-137, FIX-824, FIX-891, FIX-367, FIX-2152, FIX-2467, FIX-2371 | The coordinator | Each refresh |
+
+## Coordination seams
+
+| Seam | Between | Rule |
+|---|---|---|
+| The user line's renderer | FIX-2517, 2518 and each app | FIX-2518 owns the thumbnail and chip; apps draw sent files with it |
+| `prompt-input.tsx` | FIX-2518, 2520 | FIX-2518 lands the pieces; FIX-2520 only widens `accept` |
+| The harness turn and the board run | FIX-2519, 2522 | FIX-2519 builds both paths; FIX-2522 only writes the brief |
+| The per-target documents table | FIX-2519, 2520 | FIX-2519 lands the image rows; FIX-2520 adds the document rows |
+| Kitchen Sink's harness pages and the lab's harness path | FIX-2521, 2523 | Both consume FIX-2519; neither adapts per app |
+
+## Decided, not asked
+
+- **Limits (brief Q2).** 10 images and 5 documents a message. Images downscaled in the browser to
+ 1,568 px on the longest side, Claude's recommended size and inside every provider's limits
+ (Bedrock's 5 MB too); up to 20 MB before that. 3 MB of files a message after downscaling
+ ([Q1](#q1)). PDFs up to 100 pages. PNG, JPEG, WebP, GIF; HEIC refused. One set everywhere (ER-2).
+- **Documents per target (brief Q3).** Text-like files go as text headed by the filename, to every
+ target. A PDF goes as a file part to a generator whose provider takes PDFs, as a document block
+ to Claude Code, and as a file in the turn's directory, named in the prompt, to Codex and
+ Cursor. A provider with no PDF input gets ER-4's message; no extraction library in v1.
+- **Harness hand-off (brief Q4).** Claude Code: image and document blocks on a streaming-input
+ message. Cursor: `images` as base64. Codex: `local_image` paths. Disk files follow ER-8; if a
+ sandbox can't read outside its checkout, a directory `.git/info/exclude` hides.
+- **History (brief Q5).** A generator replays earlier turns' files, newest first, up to 10 images
+ and 5 documents; older ones become a stub naming the file, so a request holds 20 images at
+ most. A harness keeps its own transcript, so a turn sends only its own files.
+- **Steering.** A line with files sent during a turn takes `harnessSteering`'s next-turn path.
+- **Drafts.** A draft's files survive a session switch in the page, never in `localStorage`
+ (about 5 MB).
+- **A task's files are its brief's.** Settled by the owner, 2026-11-05 (brief Q7): files attached
+ while creating a task become part of its brief; the worker sees them, a re-run or hand-off
+ keeps them (ER-7).
+- **FIX-137 is not folded** (owner, 2026-11-05, brief Q6). FIX-2517 is its chat slice.
+- **Word, Excel and PowerPoint are cut** from FIX-2520 (owner, 2026-11-05), to a separate later
+ issue this epic does not file.
+- **FIX-2520's in-app PDF checks are the closure's leg b.** Its own check proves PDFs at the
+ framework and harness level, so it needn't wait for the apps.
+
+## What the end-state POC showed
+
+None built. The one premise that could break the division is the kill line, and FIX-2517's first
+PR measures it directly (ER-9); a POC would run the same measurement twice.
+
+## Analysis
+
+**2026-11-05 · watermark 2026-11-05 · right approach, with changes.** Run by hand: this session
+could not start sub-agents, so the `adhd` pass did not run. `reuse-review` folded four Adds into
+existing pieces: `usePromptInputAttachments` and `PromptInputProvider` for the hook; the
+composer's `accept`/`maxFiles`/`maxFileSize` for the limits; `FileContent` for FIX-137's item
+type; `harnessSteering`'s next-turn path for a steer with files. Per-turn files stay new, after
+the workspace host. Other frameworks: the AI SDK's chat UI sends files as data-URL parts inside
+the message (Q1's inline); hosted chat apps upload first (FIX-824). Major concern: the 4.5 MB
+limit, now Q1's cap and ER-9. Strongest other division: one issue per layer, no app issues;
+rejected, because an app's send path is where today's bug is.
+
+## How it got here
+
+- **Drafted (Nov 5).** From the product brief and its audit, verified on `main` at 9c75ce2.
+ Closure FIX-2565 filed. The owner's objective lines, the task-brief call, the FIX-137 call
+ and the Word/Excel cut are folded in. The brief's questions 1 to 5 are answered here, and two
+ of them became Q1 and Q2.
+
+**Open:** [Q1](#q1), [Q2](#q2).
diff --git a/specs/epics/FIX-2516/SPEC.md b/specs/epics/FIX-2516/SPEC.md
new file mode 100644
index 0000000000..e9a1f69519
--- /dev/null
+++ b/specs/epics/FIX-2516/SPEC.md
@@ -0,0 +1,182 @@
+# FIX-2516 · Attachments: paste or upload images and documents in every chat
+
+**Spec** · [Decisions](DECISIONS.md)
+
+## Five teams, before and after
+
+| A team that… | Today | After this epic |
+|---|---|---|
+| **chats in Kitchen Sink, Shift Manager or a lab** | Kitchen Sink drops a pasted screenshot on send without a word. The others can't attach at all | Paste, **+** or drop; see and remove each file; the reply shows the model saw it, after a reload too |
+| **runs a coding worker on Claude Code, Codex or Cursor** | Describes the bug in words: the adapters send a string | The turn's images reach the harness; a type it can't take is said, never dropped |
+| **files a task in Shift Manager with a screenshot** | The worker never sees it | The image is in the task's brief, so a re-run or hand-off keeps it |
+| **builds a flow on FSD** | Hand-rolls base64 input, limit checks and rendering ([FIX-824](https://linear.app/fixpoint-labs/issue/FIX-824)) | Sends `attachments` beside the input; the engine checks limits and writes them on the user line; a generator gets file parts |
+| **builds its own composer** (Shift Manager, the labs) | Rebuilds paste, picker and previews, or goes without | Installs the registry's pieces with `fsdev ui add` |
+
+**Why now.** The owner asked on 2026-11-03, and most of it is built and unused: the registry's
+`PromptInput` has the **+** picker, paste, drop, limits and data-URL conversion, and AI SDK 7
+takes file parts. Meanwhile Kitchen Sink silently throws a pasted image away. And the labs
+([FIX-2152](https://linear.app/fixpoint-labs/issue/FIX-2152)) exist to reproduce chat apps
+people know, whose most visible habit ours lack.
+
+## The goal, and how we'll know it's met
+
+**In every chat in our apps, a person can attach images and documents to a message (paste, a +
+picker, drag and drop), see and remove them before sending, and the reply shows the model or
+harness saw them.** The same in Kitchen Sink (Flow State chat and the Claude Code, Codex and
+Cursor pages), Shift Manager, the Claude Code-style lab and the Teammates lab.
+
+| Is it the right goal? | |
+|---|---|
+| **The real need** | People work by showing: a screenshot of the bug, the PDF of the spec. Every chat app they use takes both; ours take neither |
+| **Smaller, and rejected** | "Images reach the model in Flow State chat." FIX-2517, 2518 and 2521 meet it, and leave the coding harnesses, where a screenshot matters most, Shift Manager's tasks and both labs blind |
+| **Bigger, and not this epic's** | An upload boundary and blob store ([FIX-824](https://linear.app/fixpoint-labs/issue/FIX-824), [FIX-367](https://linear.app/fixpoint-labs/issue/FIX-367)) · upload once, reference many ([FIX-891](https://linear.app/fixpoint-labs/issue/FIX-891)) · everything under *Not doing* |
+| **Not done if** | An app on the list lacks paste, **+** or drop · a reload loses an image · a refused file is dropped without a message · a task's re-run loses its brief's images · a harness or provider drops a type without saying so · the closure check hasn't run · a bug the closure run found is open |
+
+```mermaid
+flowchart LR
+ A["each app · five chats"] --> L1["leg a · paste one image, pick a second, send"]
+ A --> L2["leg b · attach a PDF, ask what only it says"]
+ A --> L3["leg c · an 11th image, an oversized file, an unsupported type"]
+ L1 -->|"reply describes both · reload shows both"| P["PASS · the goal is met"]
+ L2 -->|"reply carries the PDF's code word"| P
+ L3 -->|"a plain message · nothing sent"| P
+ C["control · today's main"] -.-> A
+ A -.->|"under the control"| F["must FAIL · every leg"]
+```
+
+Leg c is the one a happy-path check skips; it is where Kitchen Sink fails today. Today's `main`
+must fail every leg, which is what makes the pass mean something.
+
+| How we verify | |
+|---|---|
+| **Goal check** | The closure issue, [FIX-2565](https://linear.app/fixpoint-labs/issue/FIX-2565), on one `main` commit after every other child merges, through each app's browser UI, with the real model and the real signed-in harness |
+| **Signal** | Leg a: the reply names what is in each image, and a reload shows both on the user line. Leg b: the reply contains a code word that appears only in the PDF. Leg c: the composer shows the message and the network log shows no action request |
+| **Input** | Two fresh images with distinct, checkable content and a fresh PDF with a random code word, made per run, so a model cannot guess |
+| **Anti-game** | Assert on what the person sees, never on a child's test or a stubbed model. A harness that reads the file from disk counts only when the code word reaches the reply |
+| **Control that must fail** | Today's `main`: every leg must FAIL. A planted server-side strip of `file` parts must FAIL legs a and b, which proves the reply depends on the bytes |
+
+| | |
+|---|---|
+| **Lead measure** | The children whose goal check has passed, listed by name. **None today** |
+| **Not doing** | Attachments in model replies · audio and video · a file library or reuse across sessions · cloud drive pickers · Word (.docx) and Excel (.xlsx), cut by the owner on 2026-11-05 to a separate later issue · folding [FIX-137](https://linear.app/fixpoint-labs/issue/FIX-137), which stays the broader framework issue, linked as related |
+| **Kill line** | "If data URLs inside the message can't carry 10 downscaled images under our host's request-size limit without bloating stored sessions, stop the app work and build the upload boundary (FIX-824) first." |
+
+## What's in the box
+
+
+
+Inside the box is what every chat gets once it installs the composer pieces. The fence is
+[Q1](DECISIONS.md#q1): bytes travel inside the request, so one limits set sized to the
+strictest host keeps every app the same, and the kill line says when the fence moves.
+
+**Docs:** one story, "a user turn can carry files", taught once in the guide section FIX-2517
+publishes; FIX-2519 adds the per-harness table to it and FIX-2520 the documents. FIX-2518
+publishes the registry page. Each app documents nothing of its own.
+
+## The set · as of 2026-11-05
+
+A dated snapshot. Live state is Linear and the implementation PRs. Every child is on the
+spec route and will be specced together in one spec-set PR after this one merges.
+
+| Issue | What it delivers | Why the set needs it | Status |
+|---|---|---|---|
+| [FIX-2517](https://linear.app/fixpoint-labs/issue/FIX-2517) · framework | `attachments` beside the input; refusal before any block runs; `file` parts on the user item; generator file parts and history. **Its first PR measures the kill line** | Every other child uses its shape ([ER-1](DECISIONS.md#who-owns-what)) | Backlog · spec route |
+| [FIX-2518](https://linear.app/fixpoint-labs/issue/FIX-2518) · UI library | **+**, paste, drop, previews, remove, refusals, downscaling, keyboard access; a piece own-composer apps install | Else each app rebuilds it | Backlog · spec route |
+| [FIX-2519](https://linear.app/fixpoint-labs/issue/FIX-2519) · harnesses | Images to Claude Code, Codex and Cursor; **a board-run task hands its brief's files to every run** | Where a screenshot matters most | Backlog · spec route · blocked by 2517 |
+| [FIX-2520](https://linear.app/fixpoint-labs/issue/FIX-2520) · documents, **narrowed** | PDF and text-like files (.txt .md .csv .json .yaml .html .xml, source). Word, Excel, PowerPoint cut ([decided](DECISIONS.md#decided-not-asked)) | Documents differ per target ([D1](DECISIONS.md#d1)) | Backlog · spec route · blocked by 2517, 2518, 2519 |
+| [FIX-2521](https://linear.app/fixpoint-labs/issue/FIX-2521) · Kitchen Sink | Flow State chat and three harness pages; the silent drop fixed, test first | Two of five chats | Backlog · spec route · blocked by 2517, 2518, 2519 |
+| [FIX-2522](https://linear.app/fixpoint-labs/issue/FIX-2522) · Shift Manager | Every composer and `sendTurn`; **files attached while creating a task become part of its brief** | Where a file outlives its turn | Backlog · spec route · blocked by 2517, 2518, 2519 |
+| [FIX-2523](https://linear.app/fixpoint-labs/issue/FIX-2523) · Claude Code-style lab | Its composer, through the harness path | One chat | Backlog · spec route · blocked by 2518, 2519 |
+| [FIX-2524](https://linear.app/fixpoint-labs/issue/FIX-2524) · Teammates lab | Its composer, to generator and coding teammates | One chat | Backlog · spec route · blocked by 2517, 2518, 2519 |
+| [FIX-2565](https://linear.app/fixpoint-labs/issue/FIX-2565) · closure · **required** | The QA run on one `main` commit: three legs per app, each team's journey, each child's check, the seams | Proves the whole | Backlog · blocked by every other child |
+
+Eight children and a closure, all in Backlog. Is eight really six? The labs stay apart because
+each sends to a different flow; **the collapse trigger is both lab specs having no decision of
+their own**, and then the spec set merges them. FIX-2520 stays because Codex and Cursor take no
+documents.
+
+## How the issues flow into each other
+
+```mermaid
+flowchart LR
+ X["FIX-824 · upload boundary · not in the set"] -.->|"only if the kill line trips"| A
+ A["FIX-2517 · framework"] -->|"the shape and the limits"| H["FIX-2519 · harnesses"]
+ A -->|"file parts on the user line"| APPS
+ B["FIX-2518 · UI library"] -->|"the composer pieces"| APPS
+ H -->|"a turn's files reach each harness"| APPS
+ A --> D["FIX-2520 · documents"]
+ B --> D
+ H --> D
+ subgraph APPS["the apps"]
+ KS["FIX-2521 · Kitchen Sink"]
+ SM["FIX-2522 · Shift Manager"]
+ CC["FIX-2523 · Claude Code-style lab"]
+ TM["FIX-2524 · Teammates lab"]
+ end
+ APPS --> Z["FIX-2565 · closure · required"]
+ D --> Z
+```
+
+An edge is what one issue hands the next; the dashed input is not owned. FIX-2517 and FIX-2518
+start in parallel; FIX-2523 takes FIX-2517's shape through FIX-2519. FIX-2520 runs beside the
+apps, and the closure waits on all eight.
+
+## What stays as it is
+
+- **Action input stays JSON.** No multipart boundary or blob store; that is FIX-824, which
+ enters only through the kill line.
+- **The worker door's `{ message }`** and every flow's input schema ([Q2](DECISIONS.md#q2)).
+- **Steering a running turn stays text.** A line with files waits for the next turn.
+- **Files in model replies**, handled separately.
+- **Related, not children:** FIX-137, FIX-824, FIX-891, FIX-367, FIX-2152, FIX-2467, FIX-2371.
+
+## What's already here, and the debt
+
+No retained spec designs attachments; the predecessors are Linear issues.
+
+| Already here | What the set does with it |
+|---|---|
+| `PromptInput`'s plumbing and `usePromptInputAttachments` (`packages/ui/registry/components/prompt-input.tsx`) | **Builds on.** FIX-2518 draws and exports it |
+| `FileContent` (`packages/contracts/src/items/content.ts`), never produced today | **Reuses** as the one stored part ([ER-1](DECISIONS.md#who-owns-what)) |
+| `itemToLLMMessages` (`packages/engine/src/context/history.ts`), text only | **Extends** to map `file` parts to AI SDK parts |
+| React's optimistic user line (`useSession`) | **Extends** to carry the files |
+| `HarnessRunInput { prompt }` (`packages/core/src/types/harness.ts`) | **Extends** with the turn's files |
+| `harnessSteering`'s next-turn path | **Reuses** for a line with files |
+| The task row's `goal` and `context` (`packages/orchestration/src/tasks/schema/task.ts`) | **Extends**: the brief gains files |
+| `ephemeral` content | **Leaves alone**: stripping bytes breaks the reload leg |
+| FIX-137: an `attachment` item type, request-scope storage by URL | **Amended**: this epic is its chat slice; no item type ([ER-1](DECISIONS.md#who-owns-what)); storage is [Q1](DECISIONS.md#q1)'s. It stays open |
+| FIX-824's upload boundary | **Kept** as the kill line's contingency |
+
+| Adds | Existing piece considered | Why it doesn't fit / how composed |
+|---|---|---|
+| `attachments` beside the input, checked by the server | Each action's input schema; the transcription route's byte cap | Extends the request and its parse step, where input is refused today ([Q2](DECISIONS.md#q2)). The audio cap guards another route |
+| One limits set in core | `PromptInput`'s `accept`, `maxFiles`, `maxFileSize` | Extends: the props default to core's set |
+| A registry attachment piece | `usePromptInputAttachments`, `PromptInputProvider` | Extends: lifted for composers without `PromptInput` |
+| Per-turn files for Codex and Cursor | The workspace host | New and small: the host places checkouts, not per-turn files ([ER-8](DECISIONS.md#who-owns-what)) |
+
+| The debt | |
+|---|---|
+| **Removes** | Kitchen Sink's silent drop · hand-rolled base64 chat input |
+| **Adds** | The four rows above and files on the task brief · **inline bytes in stored sessions**, paid by FIX-824 if called for |
+| **Leaves** | Each adapter's own copy of the event-to-item helper |
+
+**Net: up, on purpose.** A capability every chat app has; the named debt is inline bytes.
+
+## Sign off
+
+**[The goal](#the-goal-and-how-well-know-its-met), at that size:** five chats, images and PDFs,
+refusals that say why, proved in the browser. If wrong: we ship it in one app and the coding
+harnesses stay blind, or we carry a closure leg nobody needed.
+
+**[The set's debt](#whats-already-here-and-the-debt), net up:** one capability, and inline bytes
+in stored sessions until FIX-824. If wrong: sessions grow until a reload is slow, which the kill
+line exists to catch first.
+
+1. **[D1](DECISIONS.md#d1) · Eight children and a closure: images first, documents after.** If
+ wrong: two lab issues that should have been one.
+
+**Open: 2.**
+[Q1](DECISIONS.md#q1) · do stored sessions keep the bytes inline in v1? I recommend yes.
+[Q2](DECISIONS.md#q2) · do attachments ride beside the input in the public API? I recommend yes.
+
+Epic · 8 children and a closure · Workforce: Shift Manager · Goal 1, validate through real usage
diff --git a/specs/epics/FIX-2516/figures/d1-division.svg b/specs/epics/FIX-2516/figures/d1-division.svg
new file mode 100644
index 0000000000..0ad9cce884
--- /dev/null
+++ b/specs/epics/FIX-2516/figures/d1-division.svg
@@ -0,0 +1,100 @@
+
diff --git a/specs/epics/FIX-2516/figures/end-state.svg b/specs/epics/FIX-2516/figures/end-state.svg
new file mode 100644
index 0000000000..0df39283b9
--- /dev/null
+++ b/specs/epics/FIX-2516/figures/end-state.svg
@@ -0,0 +1,92 @@
+
diff --git a/specs/epics/FIX-2516/figures/open-inline-bytes.svg b/specs/epics/FIX-2516/figures/open-inline-bytes.svg
new file mode 100644
index 0000000000..ae8810cfcd
--- /dev/null
+++ b/specs/epics/FIX-2516/figures/open-inline-bytes.svg
@@ -0,0 +1,112 @@
+
diff --git a/specs/epics/FIX-2516/figures/open-request-field.svg b/specs/epics/FIX-2516/figures/open-request-field.svg
new file mode 100644
index 0000000000..5f95add53c
--- /dev/null
+++ b/specs/epics/FIX-2516/figures/open-request-field.svg
@@ -0,0 +1,99 @@
+
diff --git a/specs/epics/FIX-2516/figures/ownership.svg b/specs/epics/FIX-2516/figures/ownership.svg
new file mode 100644
index 0000000000..23f22073ce
--- /dev/null
+++ b/specs/epics/FIX-2516/figures/ownership.svg
@@ -0,0 +1,185 @@
+