From 5132f36601e670b942161addee54b324993d76b3 Mon Sep 17 00:00:00 2001 From: Claude Date: Tue, 29 Sep 2026 12:14:33 +0000 Subject: [PATCH 1/3] spec(epic): hard gates before public release (FIX-1635) The retained epic set for FIX-1635: the goal (the packed release installs, imports and holds its security boundaries over real HTTP), the locked child set with routes, three cross-cutting decisions (one release-shaped proof, reproduce-first, FIX-1018 lands first), rules with owners, the path, the shared docs narrative and cross-child lineage. Closure issue FIX-1636 filed. Co-Authored-By: Claude Opus 5.5 (1M context) Claude-Session: https://claude.ai/code/session_01HNTQkGXCUryuVD2TEBLi6G --- specs/epics/FIX-1635/BUSINESS-RULES.md | 47 ++++++ specs/epics/FIX-1635/DECISIONS.md | 97 +++++++++++++ specs/epics/FIX-1635/DOCS.md | 45 ++++++ specs/epics/FIX-1635/EVOLUTION.md | 16 +++ specs/epics/FIX-1635/PLAN.md | 75 ++++++++++ specs/epics/FIX-1635/SPEC.md | 136 ++++++++++++++++++ specs/epics/FIX-1635/figures/d1-one-proof.svg | 100 +++++++++++++ .../FIX-1635/figures/d2-reproduce-first.svg | 100 +++++++++++++ .../FIX-1635/figures/d3-id-owner-first.svg | 112 +++++++++++++++ specs/epics/FIX-1635/figures/end-state.svg | 92 ++++++++++++ specs/epics/FIX-1635/figures/ownership.svg | 125 ++++++++++++++++ specs/epics/FIX-1635/figures/path.svg | 106 ++++++++++++++ 12 files changed, 1051 insertions(+) create mode 100644 specs/epics/FIX-1635/BUSINESS-RULES.md create mode 100644 specs/epics/FIX-1635/DECISIONS.md create mode 100644 specs/epics/FIX-1635/DOCS.md create mode 100644 specs/epics/FIX-1635/EVOLUTION.md create mode 100644 specs/epics/FIX-1635/PLAN.md create mode 100644 specs/epics/FIX-1635/SPEC.md create mode 100644 specs/epics/FIX-1635/figures/d1-one-proof.svg create mode 100644 specs/epics/FIX-1635/figures/d2-reproduce-first.svg create mode 100644 specs/epics/FIX-1635/figures/d3-id-owner-first.svg create mode 100644 specs/epics/FIX-1635/figures/end-state.svg create mode 100644 specs/epics/FIX-1635/figures/ownership.svg create mode 100644 specs/epics/FIX-1635/figures/path.svg diff --git a/specs/epics/FIX-1635/BUSINESS-RULES.md b/specs/epics/FIX-1635/BUSINESS-RULES.md new file mode 100644 index 0000000000..9fa27aaf0b --- /dev/null +++ b/specs/epics/FIX-1635/BUSINESS-RULES.md @@ -0,0 +1,47 @@ +# FIX-1635 · Rules every issue in the set obeys + +[Spec](SPEC.md) · [Decisions](DECISIONS.md) · **Rules** · [Plan](PLAN.md) · [Docs](DOCS.md) · [Evolution](EVOLUTION.md) + +Each rule says who owns it and where it's checked. "Closure" means FIX-1636's goal check, +run over real HTTP against installed tarballs. + +## What a team gets, and what it doesn't + +| # | Rule | Owner | Checked at | +|---|---|---|---| +| ER-1 | Every publishable package, packed and installed into an empty ESM project, imports | FIX-1431 | Closure leg a · CI's extension check | +| ER-2 | DevTool cannot be published without its client assets, and serves them from an installed copy | FIX-1334 | Closure leg a · the release build | +| ER-3 | A request id a caller supplies never reaches another user's request record or its items, on any path that writes one | FIX-1018 | Closure leg b · host and SQLite tests | +| ER-4 | Two users in one tenant with the same request id never share a run-scoped workspace or its resources | FIX-1286 (consumes ER-3) | Closure leg b | +| ER-5 | A session id reaches only its owner's session record; another user's id is not found | FIX-1022 | Closure leg b | +| ER-6 | Only an allow-listed source is re-entered from public HTTP by retry, continue or resume | FIX-1021 | Closure leg b | +| ER-7 | A session's stored flow never authorizes reading another flow's requests or items | FIX-1046 | Closure leg b | +| ER-8 | A cross-flow dispatch admits the target exactly as ingress would, organization included | FIX-1328 | Closure leg b | +| ER-9 | A schema-invalid resource write throws and leaves state as it was | FIX-1256 (shipped) | Closure re-runs its test | +| ER-10 | `writable: false` on a collection refuses every write-shaped path, `create` with `replace` and `delete` included | FIX-1261 with FIX-1510 (shipped) | Closure re-runs its tests | +| ER-11 | On a queue host, a delivery into an existing session runs the recipient under its concurrency policy and incarnation guard; what still can't be honoured is refused by name | FIX-1634 | Closure leg c | +| ER-12 | A non-streaming multi-step turn returns the text the streamed turn would | FIX-1628 | FIX-1628's test · closure re-run | + +## What no child may do + +| # | Rule | Because | +|---|---|---| +| ER-13 | No Workforce (Layer 2) concept in core, engine, client or react. FIX-1634 is generic; its Workforce adopt child is filed after its Layer 1 shape merges | The owner's layer rule | +| ER-14 | No child reopens org-optional, or deletes a named refusal without restating the guarantee it protected | FIX-1442 is settled; a deleted refusal trades a loud failure for a silent one | +| ER-15 | No child widens: Tier B leftovers, first-hour DX and kitchen-sink polish stay out. Something noticed outside the goal is filed normally, not under this epic | The set is locked by the owner | +| ER-16 | No child pins today's defect as a contract in a test, and every fix lands a regression test that fails on the commit before the hole closed | A test that can't fail proves nothing | + +## How the set is run + +| # | Rule | Because | +|---|---|---| +| ER-17 | A direct-route worker reproduces on current `main` first. No reproduction: it closes with a regression test (ER-16) and the commit that closed the hole, not a rewrite | D2 | +| ER-18 | FIX-1018 merges before FIX-1286 and FIX-1634 implement; their specs may start now | D3 | +| ER-19 | A child's Linear state is mirrored the moment it changes; its route comes from its label | The wake derives blocked-by and route from Linear | + +## The closure + +| # | The epic is done when | Proved by | +|---|---|---| +| ER-20 | [The goal](SPEC.md#the-goal-and-how-well-know-its-met) is met: legs a, b and c pass on one `main` commit, each fails under its control, and every finding is fixed and retested | FIX-1636's goal check | +| ER-21 | The authentication guide says a caller-chosen id is an address, not an ownership | FIX-1018's docs change ([DOCS.md](DOCS.md)) | diff --git a/specs/epics/FIX-1635/DECISIONS.md b/specs/epics/FIX-1635/DECISIONS.md new file mode 100644 index 0000000000..1af2c96212 --- /dev/null +++ b/specs/epics/FIX-1635/DECISIONS.md @@ -0,0 +1,97 @@ +# FIX-1635 · Decisions + +[Spec](SPEC.md) · **Decisions** · [Rules](BUSINESS-RULES.md) · [Plan](PLAN.md) · [Docs](DOCS.md) · [Evolution](EVOLUTION.md) + +The calls that sit above any single child. The child set itself is not one of them: the owner +locked it with the FSD Architect on 2026-09-29, including what is out of the epic. Two cards +are the sign-off; the third is an engineering call recorded so no child reopens it. + +## The tree + +```mermaid +flowchart TD + E["FIX-1635"] --> D1["D1 · one epic, one release-shaped proof"] + D1 -.->|"rejected"| X1["twelve independent fixes, each proved from source"] + E --> D2["D2 · reproduce first, close with a test"] + D2 -.->|"rejected"| X2["implement each ticket as written"] + E --> D3["D3 · FIX-1018 first, the id bound to its owner"] + D3 -.->|"rejected"| X3["FIX-1286 decides first, or re-key the run workspace"] +``` + + +## D1 · These twelve are one epic with one release-shaped proof + +| | | +|---|---| +| **Instead of** | Fixing each as ordinary backlog, each proved by its own test from source | +| **Because** | What blocks a release is what a consumer meets, and no test in the repo runs as one: 0.1.1 passed `publint`, a dry-run publish and every source test, then failed on its first import. Each security fix is likewise proved one function at a time, never as two users over HTTP against an installed server | +| **Locks in** | A closure run that packs, installs, serves over HTTP and needs a real Redis. It is the most expensive child and it repeats per finding | + +**What would change my mind:** evidence that the release is not going to external users soon. +Then these are ordinary bugs, and the closure harness can wait for the launch that needs it. + +![D1, is this one epic with one proof: one epic with a release-shaped closure run, chosen, beside twelve independent fixes proved from source. Decides it: a package that fails on import passed every source test. Price: a closure run that packs, installs and needs Redis. Locks in the most expensive child; flips if the release is not going to external users soon.](figures/d1-one-proof.svg) + +It comes down to 0.1.1: every source test passed and the first import failed. + + +## D2 · Reproduce on current `main` first; a hole already closed closes with a test + +| | | +|---|---| +| **Instead of** | Implementing each ticket's fix as written | +| **Because** | Most tickets were written in August against code that has moved since: the org-required pass (FIX-1442), the public re-entry allow-list (FIX-999), the ESM-extension build step and the release asset build. A fix for a hole that is gone rewrites working code | +| **Locks in** | A direct-route child that no longer reproduces still lands a PR: a regression test through the ticket's own path that fails on the commit before the hole closed. Closing with a comment alone is not done | + +![D2, what a worker does with an old ticket: reproduce on current main and close with a test, chosen, beside implementing the fix as written. Decides it: tickets written in August against code that has since moved. Price: a PR even when no code changes. Locks in a regression test for every hole; flips if the tickets were fresh.](figures/d2-reproduce-first.svg) + +It comes down to age: an August ticket describes code that later passes may have fixed. + + +## D3 · FIX-1018 lands first; a caller-supplied request id is bound to its owner + +| | | +|---|---| +| **Instead of** | FIX-1286 deciding the principal question first in a spec · or giving the run workspace its own user key | +| **Because** | FIX-1018 is a confirmed cross-user read and should not wait on a design review. Its fix answers the question FIX-1286 asks: once no user can adopt another user's request record, two users with one request id already have two records, and the workspace key may need no change. FIX-1634 extends the queue branch where FIX-1018's unguarded write lives | +| **Locks in** | FIX-1286 and FIX-1634 implement after FIX-1018 merges; both specs can start now. A caller-supplied request id stays accepted (callers use it for idempotency), scoped to its owner. FIX-1286 may close as proved by FIX-1018, with its own test | + +**What would change my mind:** FIX-1018's reproduction showing the shared workspace survives +per-owner request records. Then FIX-1286 decides independently and the edge comes out. + +![D3, which principal fix goes first: FIX-1018 binds the request id to its owner first, chosen, beside FIX-1286 deciding first in a spec and beside a user key on the run workspace. Decides it: FIX-1018 is a confirmed disclosure and its fix answers FIX-1286's question. Price: FIX-1286 and FIX-1634 wait on it. Flips if the shared workspace survives per-owner records.](figures/d3-id-owner-first.svg) + +It comes down to a confirmed leak: the design question should not hold the fix for it. + +## Who owns what + +![Who owns what: fourteen rules by the open children and the closure issue. Each security, integrity, packaging, queue and generator rule has one child that builds it; FIX-1286 and FIX-1634 consume FIX-1018's binding; the closure issue builds the proof and consumes every other rule.](figures/ownership.svg) + +Every rule has one owner. A *consumes* cell is a place a child must not re-decide: FIX-1286 +and FIX-1634 build on FIX-1018's binding rather than inventing their own. + +## Decided in review, recorded so no child reopens them + +- **Routes.** Direct: FIX-1018, 1021, 1022, 1046, 1328, 1431, 1334, 1628. Spec: FIX-1634 (how + concurrency is arbitrated across processes is a design question) and FIX-1286 (its issue + asks for a decision). The closure issue is spec: its spec is the QA plan. +- **FIX-1328's "always require an org" option is already the product's direction** + (FIX-1442). Its worker reproduces against that, and does not re-open seam-side admission as + a product fork. +- **FIX-1634 is Layer 1 only.** Nothing in it names a channel, seat or specialist. The + Workforce adopt child is filed after its Layer 1 shape merges, not before. +- **FIX-1628 stays scoped** to the non-streaming step loop. +- **The done rows are re-verified, not reopened.** The closure plan re-runs their tests; a + failure is a new bug child. + +## What the end-state POC showed + +None built. The children are independent fixes with one ordering edge, and the closure run +is the whole-set proof. + +## How it got here + +- **Drafted (Sep 29)** from the owner's locked set and the Architect's clusters; the closure + issue FIX-1636 filed; FIX-1018 wired to block FIX-1286 and FIX-1634. + +**Open: none.** diff --git a/specs/epics/FIX-1635/DOCS.md b/specs/epics/FIX-1635/DOCS.md new file mode 100644 index 0000000000..bb25f0e842 --- /dev/null +++ b/specs/epics/FIX-1635/DOCS.md @@ -0,0 +1,45 @@ +# FIX-1635 · Documentation draft + +[Spec](SPEC.md) · [Decisions](DECISIONS.md) · [Rules](BUSINESS-RULES.md) · [Plan](PLAN.md) · **Docs** · [Evolution](EVOLUTION.md) + +One shared story changes: which ids a caller may choose, and what choosing one gives them. +Most children change behaviour the docs already promise (a package imports, a guard refuses), +so they publish nothing new. The specifics below belong to the child that ships them. + +## UPDATE · `apps/docs/docs/server/authentication.md` · new subsection after "Addressed routes, and what they scope by" + +> ### Ids you choose are addresses +> +> A client can pick its own ids. An action call may carry a `requestId`, and most routes name +> a session by id. That is useful: a retried call with the same `requestId` is recognised as +> the same request instead of starting a second one. +> +> An id is never proof of ownership. A request or a session belongs to the user and +> organization it was created under, and an id someone else created reaches nothing of yours. +> Send another user's `requestId` and you get your own request, not theirs. Send another +> user's session id and you get a `404`, the same answer as an id that doesn't exist. +> +> ``` +> POST /api/flows/chat/actions/send bob's, requestId req_1 (alice's) -> bob's own request +> GET /api/flows/sessions/s_alice bob's, same tenant -> 404 +> ``` +> +> So you don't need to make ids unguessable to keep users apart. Ids still leak in headers, +> URLs and logs, and nothing here relies on them staying secret. + +The example routes are illustrative until FIX-1018's `DOCS.md` fixes the exact paths and +status codes against the shipped behaviour. + +## Ownership + +| Material | Publisher | Specific draft | +|---|---|---| +| "Ids you choose are addresses", above | FIX-1018, once its behaviour is verified | This document | +| The session-id half of the example, and legacy session keys | FIX-1022 | Its PR; links here | +| A session's flow does not authorize another flow's history | FIX-1046 | Its PR; one sentence in "Addressed routes" | +| The `external-dispatcher` row in `server/background-work.md`'s refusal table, narrowed to what a queue host still can't honour | FIX-1634 | Its `DOCS.md` | +| `writable` on collections | Shipped with FIX-1510 | None | + +FIX-1021, 1328, 1431, 1334, 1286 and 1628 restore behaviour the docs already describe, so they +publish nothing. If a worker finds a page that promised the broken behaviour, it corrects that +page in the same PR. No unchanged page is copied here. diff --git a/specs/epics/FIX-1635/EVOLUTION.md b/specs/epics/FIX-1635/EVOLUTION.md new file mode 100644 index 0000000000..6a53f99a66 --- /dev/null +++ b/specs/epics/FIX-1635/EVOLUTION.md @@ -0,0 +1,16 @@ +# FIX-1635 · Evolution + +[Spec](SPEC.md) · [Decisions](DECISIONS.md) · [Rules](BUSINESS-RULES.md) · [Plan](PLAN.md) · [Docs](DOCS.md) · **Evolution** + +Only lineage that spans more than one child. Child-specific lineage belongs in FIX-1634's and +FIX-1286's own specs. + +| Prior intent and precise source | Treatment | Reason / evidence | Replacement | Compatibility | +|---|---|---|---|---| +| FIX-1442: every session and request carries an organization, owned by the verified principal. Source: its Linear spec document and review PR #1899 (no retained repository set) | **Retained** | Shipped (Done, 2026-09). FIX-1328's second option is this rule; the principal cluster builds on it | ER-8, ER-14 | Nothing changes for org identity | +| FIX-682: session keys and re-entry scoped by tenant. Source: `packages/engine/src/stores/scope-keys.ts` → `resolveSessionStorageKey` | **Amended** by FIX-1022 | The tenant wall holds; users inside one tenant were never separated | ER-5 | Existing tenant-only keys are still read by their owner (BP-030) | +| FIX-999: public re-entry admits an allow-list of sources. Source: `packages/engine/src/routes/public-reentry.ts` | **Retained**; FIX-1021 verifies against it | The allow-list landed after FIX-1021 was written | ER-6 | None | +| FIX-1302: delivery into an existing session is refused `external-dispatcher` on a queue host. Source: `docs/architecture/inbound-transports.md`, the `external-dispatcher` paragraph | **Amended** by FIX-1634 | The refusal protected recipient concurrency and the incarnation guard; FIX-1634 restores both across the queue boundary instead of deleting it | ER-11, ER-14 | Anything still unhonourable keeps a named refusal | + +Nothing is wholly superseded. Re-check each cited intent against current code before +implementing (ER-17). diff --git a/specs/epics/FIX-1635/PLAN.md b/specs/epics/FIX-1635/PLAN.md new file mode 100644 index 0000000000..18ea9ee791 --- /dev/null +++ b/specs/epics/FIX-1635/PLAN.md @@ -0,0 +1,75 @@ +# FIX-1635 · Plan + +[Spec](SPEC.md) · [Decisions](DECISIONS.md) · [Rules](BUSINESS-RULES.md) · **Plan** · [Docs](DOCS.md) · [Evolution](EVOLUTION.md) + +An epic plan sequences the work and says what each piece entails. It does not say how to build +any piece. IDs cross-reference [DECISIONS.md](DECISIONS.md) (D-n) and +[BUSINESS-RULES.md](BUSINESS-RULES.md) (ER-n). + +## The path + +![The path as of 29 September 2026, in steps rather than dates because the epic is unscheduled. Shipped: FIX-1256, FIX-1261 and FIX-1510. Now: this spec in review. Step 1, in parallel: FIX-1018; the principal bugs FIX-1022, 1021, 1046 and 1328; the release bugs FIX-1431 and 1334; FIX-1628; the specs for FIX-1286 and FIX-1634; the closure QA plan. Step 2: FIX-1286 and FIX-1634 implement after FIX-1018 merges. Step 3: the closure run on one main commit, its PR, then wrap. The critical path runs through FIX-1018, FIX-1634's implementation and the closure run.](figures/path.svg) + +A wide first step and one chain. Everything except two implementations can start at the +gate; the critical path is FIX-1018, then FIX-1634, the largest child, then the closure run. +The dependency graph is in [the spec](SPEC.md#how-the-issues-flow-into-each-other). + +## What each issue entails + +| Issue | Route | Consumes | Delivers | Releases | Size | +|---|---|---|---|---|---| +| **FIX-1018** request id | direct | Current `main` (ER-17) | ER-3; the docs line (ER-21) | FIX-1286 · FIX-1634 | Medium | +| **FIX-1286** run workspace | spec → impl | ER-3 | ER-4, or a test proving ER-3 already delivers it | The closure | Small, or none | +| **FIX-1022** session key | direct | Current `main` | ER-5, with legacy keys still read by their owner | The closure | Medium | +| **FIX-1021** re-entry | direct | Current `main` | ER-6 | The closure | Small, likely a test | +| **FIX-1046** sibling flow | direct | Current `main` | ER-7 | The closure | Small | +| **FIX-1328** cross-flow admission | direct | FIX-1442's org rule | ER-8 | The closure | Small, likely a test | +| **FIX-1431** `dist` extensions | direct | Current `main` | ER-1 | The closure | Small, likely a test | +| **FIX-1334** DevTool assets | direct | Current `main` | ER-2 | The closure | Small, likely a test | +| **FIX-1634** queue delivery | spec → impl | ER-3 · FIX-1302's refusal | ER-11, Layer 1 only | The closure; later, a Workforce adopt child | Large | +| **FIX-1628** non-streaming text | direct | FIX-1626's join rule | ER-12 | The closure | Small | +| **FIX-1636** closure · required | spec (the QA plan) → runs until clean → PR | Every other child, merged, on one `main` commit | The committed release check, a QA report, a bug child per finding | Wrap | Large, repeats per retest | + +"Likely a test" is not a diagnosis. Each worker's **first step is reproducing the ticket's hole +on current `main`** (ER-17): FIX-1442, FIX-999's re-entry allow-list, the ESM-extension build +step and the release asset build all landed after these tickets were written. + +## Where it is + +[The set table](SPEC.md#the-set--as-of-2026-09-29) is the review-time snapshot; follow its +Linear links for live state. Inputs from outside the set, all shipped: FIX-1442 (org required), +FIX-1302 (the `external-dispatcher` refusal FIX-1634 narrows), FIX-1626 (the streamed join +FIX-1628 matches). + +## What unblocks what, from here + +1. **This spec merges** → every child's spec or reproduction starts; the closure QA plan is + written while they build. +2. **FIX-1018 merges** → FIX-1286 and FIX-1634 implement against its binding (D3). +3. **Every other child merges** → the closure runs on one `main` commit. +4. **A run files findings** → each is a child of FIX-1635 that blocks FIX-1636; the whole plan + re-runs on a fresh commit when the last merges. +5. **A run files nothing** → the closure PR opens with the checks and the QA report. + +## Coordination seams to watch + +| Seam | Between | Rule | +|---|---|---| +| The request write on the queue and external-dispatcher branches (`createInboundTransportHost`) | FIX-1018 and FIX-1634 | 1018 guards it first; 1634 delivers through it and keeps the guard | +| Session adoption checks (`createExecutionContext`) | FIX-1022 and FIX-1046 | Each adds its own binding beside user, tenant and org; whichever lands second rebases, never merges the two | +| Session and request routes' owner checks | FIX-1022, FIX-1046, FIX-1021 | Not-found for another user's record, one shape across all three | +| The run-scope key | FIX-1018 and FIX-1286 | 1286 changes it only if 1018's binding leaves a shared workspace | +| `server/authentication.md` | FIX-1018, FIX-1022, FIX-1046 | 1018 publishes the id rule; the others link to it | +| The publish path (`release:build`, CI's `dist` check) | FIX-1431 and FIX-1334 | One release build; neither adds a second | + +## Not children, deliberately + +The Tier B leftovers (FIX-1002 family and the owner's list), FIX-1591, FIX-1631 to 1633, +FIX-1621, FIX-1161's publish and ops work, and the Workforce adopt child for FIX-1634 until +its Layer 1 shape merges. + +## Wrap + +When ER-20 holds: the lessons pass, a docs polish over the authentication and background-work +pages, and a completion report in Linear. The original merged spec PR stays as the review +record. diff --git a/specs/epics/FIX-1635/SPEC.md b/specs/epics/FIX-1635/SPEC.md new file mode 100644 index 0000000000..e080dd8fd5 --- /dev/null +++ b/specs/epics/FIX-1635/SPEC.md @@ -0,0 +1,136 @@ +# FIX-1635 · Hard gates before public release: the release installs, and holds its boundaries + +**Spec** · [Decisions](DECISIONS.md) · [Rules](BUSINESS-RULES.md) · [Plan](PLAN.md) · [Docs](DOCS.md) · [Evolution](EVOLUTION.md) + +Epic · 12 children and a closure issue · Public Launch · Goal 4, keep the foundation honest, +with Goal 3 ([`docs/objectives.md`](../../../docs/objectives.md)) · +[FIX-1635](https://linear.app/fixpoint-labs/issue/FIX-1635) · child set locked by the owner +with the FSD Architect, 2026-09-29 + +## Five teams, before and after + +| A team that… | Today | After this epic | +|---|---|---| +| **installs FSD from npm** | 0.1.1 failed on its first import; nothing in CI installs a packed tarball and imports it | Every published package imports from a clean install, and DevTool ships its client assets | +| **puts several users in one tenant** | A caller who knows another user's request or session id can reach that record, its items or its sandbox | An id a caller chooses is an address, never an ownership. Another user's id is not found | +| **exposes flows over public HTTP** | Re-entry, sibling-flow listings and cross-flow dispatch each trust something the caller controls | Each admits on the stored record and the target flow's own rules | +| **runs work on a queue host (BullMQ)** | Delivery into an existing session is refused, so channels and escalations do nothing there | The recipient runs under its own concurrency policy; what still can't be honoured is refused by name | +| **calls a generator without streaming** | A multi-step turn returns only its last step's text | The same text as the streamed turn | + +**Why now.** The release is the first time someone outside this repository runs FSD, and each +of these is a defect they would meet, not a missing feature. A leaked record or a package that +won't import is found by the first user, and costs trust before any feature is judged. + +## The goal, and how we'll know it's met + +**Someone who installs FSD from the packed release can import every package and run a server, +and none of the holes this epic names can be reached over real HTTP.** + +| Is it the right goal? | | +|---|---| +| **The real need** | The owner's cut, 2026-09-29: the defects that block a public release, not polish, not parked bets, not first-hour DX | +| **Smaller, and rejected** | "Every child is Done." Each child's test runs from source, and a package that fails on import passed every source test in the repo | +| **Bigger, and not this epic's** | First-hour DX ([FIX-1161](https://linear.app/fixpoint-labs/issue/FIX-1161)) · the Tier B leftovers · Workforce adopting the queue delivery · kitchen-sink polish | +| **Not done if** | A leg runs against source instead of installed tarballs · a security leg calls a store or function directly instead of HTTP · a hole closed as "no longer reproduces" has no test that fails on the commit before it closed · the checks ran on different commits · a closure finding is open | + +```mermaid +flowchart LR + A["pack every package · install into an empty project"] --> L1["leg a · import each, start a server"] + L1 --> L2["leg b · two users, one tenant, one leg per hole over HTTP"] + L1 --> L3["leg c · BullMQ host with real Redis"] + L2 --> R["what each caller gets back"] + L3 --> R + R -->|"every leg holds"| P["PASS · the epic's goal is met"] + C["control · last npm release, or the commit before each fix"] -.-> R + R -.->|"under the control"| F["must FAIL · names its leg"] +``` + +Leg a is the one no child's own test reaches. Leg b is where each hole is asked for as an +attacker would ask for it. + +| How we verify | | +|---|---| +| **Goal check** | [FIX-1636](https://linear.app/fixpoint-labs/issue/FIX-1636)'s QA plan, on one `main` commit after every other child merges ([ER-20](BUSINESS-RULES.md#the-closure)) | +| **Model** | Scripted and keyless. The generator leg asserts on text a scripted model wrote, not on a model's quality | +| **Signal** | **a:** every packed package imports in an empty ESM project, and DevTool serves its client assets from the installed copy. **b:** as user B in user A's tenant, each named hole returns not-found or refused, and A's records are unchanged afterwards. **c:** a delivery into an existing session runs the recipient on a queue-backed host | +| **Input** | Ids B learns the way an attacker would: A's request id from a response header, A's session id from a URL | +| **Anti-game** | No leg imports from `src`, calls a store or engine function directly, or runs against a workspace link. Each passes while a consumer still breaks | +| **Control that must fail** | Leg a against the last published npm version. Each leg b and c case against the commit just before its fix merged, including fixes that shipped before this epic ([D2](DECISIONS.md#d2)) | + +## What's in the box + +![What's in the box: the packed release imports and DevTool carries its assets; ids chosen by a caller never reach another user's records; public entry points admit on the stored record; resource guards throw or refuse; a queue host delivers into an existing session; a non-streaming turn keeps its text. Fence: no Workforce word below Workforce, no reopening org-optional, no refusal deleted without its guarantee. Composed in by the app: its principal resolver, its tenant header, its queue host. Not built: first-hour DX, the Tier B leftovers, Workforce adopting queue delivery, kitchen-sink polish.](figures/end-state.svg) + +Inside the box is what any app gets from the release without writing a line. The fence is +what keeps a hard-gates epic from turning into a redesign. + +## The set · as of 2026-09-29 + +A dated snapshot. Live state is Linear and the implementation PRs. Direct-route rows have no +spec PR by design. + +| Issue | What it delivers | Why the set needs it | Route · status | +|---|---|---|---| +| [FIX-1256](https://linear.app/fixpoint-labs/issue/FIX-1256) · invalid resource write | Throws instead of resetting state | Silent data loss | direct · **Done** · [#1469](https://github.com/fixpoint-labs/flow-state-dev/pull/1469) | +| [FIX-1261](https://linear.app/fixpoint-labs/issue/FIX-1261), remainder [FIX-1510](https://linear.app/fixpoint-labs/issue/FIX-1510) · `writable: false` on collections | Refused on every write-shaped path | A guard a neighbour bypasses | direct · **Done** · [#1470](https://github.com/fixpoint-labs/flow-state-dev/pull/1470), [#2047](https://github.com/fixpoint-labs/flow-state-dev/pull/2047) | +| [FIX-1018](https://linear.app/fixpoint-labs/issue/FIX-1018) · request id reparenting | A caller's request id cannot reach another user's record or items | Confirmed cross-user read; first of the principal cluster ([D3](DECISIONS.md#d3)) | direct · Backlog | +| [FIX-1286](https://linear.app/fixpoint-labs/issue/FIX-1286) · run-scoped workspace | Two users with one request id never share a sandbox | Its issue asks for a decision, not a patch | spec · Backlog · blocked by 1018 | +| [FIX-1022](https://linear.app/fixpoint-labs/issue/FIX-1022) · session key has no principal | Another user's session id is not found | Same-tenant session read | direct · Todo | +| [FIX-1021](https://linear.app/fixpoint-labs/issue/FIX-1021) · re-entry deny-lists | Only allow-listed sources re-enter from HTTP | Caller-chosen input under another source's privileges | direct · Todo | +| [FIX-1046](https://linear.app/fixpoint-labs/issue/FIX-1046) · sibling-flow listing | A session's stored flow never authorizes another flow's history | Protected history readable | direct · Backlog | +| [FIX-1328](https://linear.app/fixpoint-labs/issue/FIX-1328) · cross-flow admission | The seam admits as ingress does | A target runs without the org it requires | direct · Backlog | +| [FIX-1431](https://linear.app/fixpoint-labs/issue/FIX-1431) · extensionless `dist` | Published packages import | The release is unusable without it | direct · Backlog | +| [FIX-1334](https://linear.app/fixpoint-labs/issue/FIX-1334) · DevTool client assets | DevTool can't publish without them | A first publish breaks at runtime | direct · Backlog | +| [FIX-1634](https://linear.app/fixpoint-labs/issue/FIX-1634) · queue-host delivery | Delivery into an existing session on BullMQ, Layer 1 only | Channels and escalations do nothing on a queue host | spec · Backlog · blocked by 1018 | +| [FIX-1628](https://linear.app/fixpoint-labs/issue/FIX-1628) · non-streaming text drop | A multi-step turn keeps every step's text | Same turn, two answers | direct · Backlog | +| [FIX-1636](https://linear.app/fixpoint-labs/issue/FIX-1636) · closure · **required** | The QA plan, run on one `main` commit | The only child that runs the release as a consumer does | spec · Backlog · blocked by every open child | + +**3 done · 10 open · the closure.** Could it be smaller? The set is the owner's, locked with +the Architect, and this spec does not reopen it. What can shrink is the work: several tickets +predate later passes on `main`, so each direct-route worker reproduces first, and a hole that +is already closed closes with a test instead of a rewrite ([D2](DECISIONS.md#d2)). + +## How the issues flow into each other + +```mermaid +flowchart LR + S["shipped · FIX-1256, 1261, 1510"] --> Z["FIX-1636 · closure · required"] + A["FIX-1018 · request id bound to its owner"] -->|"the binding"| B["FIX-1286 · run workspace"] + A -->|"a guarded request write"| Q["FIX-1634 · queue-host delivery"] + P["FIX-1022, 1021, 1046, 1328 · principal holes"] --> Z + K["FIX-1431, 1334 · the release loads"] --> Z + G["FIX-1628 · non-streaming text"] --> Z + A --> Z + B --> Z + Q --> Z + classDef done stroke-width:2px + class S done +``` + +Edges are blocked-by on implementation; every spec can start at the gate. Only FIX-1018 gates +other children. The clusters are a sequencing hint, not dependencies. + +## What stays as it is + +- **Org is never optional** ([FIX-1442](https://linear.app/fixpoint-labs/issue/FIX-1442)). No child reopens it. +- **Cross-tenant walls** (FIX-682). The holes here are inside one tenant. +- **Workforce.** Adopting queue delivery is a later child, filed once FIX-1634's Layer 1 shape merges. +- **Out of this epic, by the owner:** the Tier B leftovers, [FIX-1591](https://linear.app/fixpoint-labs/issue/FIX-1591), kitchen-sink polish (FIX-1631 to 1633), [FIX-1621](https://linear.app/fixpoint-labs/issue/FIX-1621), and publish work under [FIX-1161](https://linear.app/fixpoint-labs/issue/FIX-1161). + +## Sign off + +**[The goal](#the-goal-and-how-well-know-its-met), at that size:** the packed release, +installed and served, with every named hole asked for over HTTP. If wrong: we launch on source +tests that a consumer's install never ran, or carry a closure harness heavier than the fixes. + +1. **[D1](DECISIONS.md#d1) · These twelve are one epic with one release-shaped proof.** If + wrong: we spend a closure run on fixes that were each fine alone. +2. **[D3](DECISIONS.md#d3) · FIX-1018 lands first; a caller's id is bound to its owner, and + FIX-1286 specs against that.** If wrong: FIX-1286 and FIX-1634 wait a few days on a fix they + didn't need. + +**Recorded, not asked:** [D2](DECISIONS.md#d2), reproduce first and close with a test. + +**Open: none.** Approve this direction by merging. Reasoning and what lost: +[DECISIONS.md](DECISIONS.md). Rules: [BUSINESS-RULES.md](BUSINESS-RULES.md). Order: +[PLAN.md](PLAN.md). diff --git a/specs/epics/FIX-1635/figures/d1-one-proof.svg b/specs/epics/FIX-1635/figures/d1-one-proof.svg new file mode 100644 index 0000000000..a2ca2ea3b8 --- /dev/null +++ b/specs/epics/FIX-1635/figures/d1-one-proof.svg @@ -0,0 +1,100 @@ + + + + + + + D1 · THE CALL + Are these twelve one epic with one release-shaped proof? + + + + + CHOSEN + One epic, one closure run + + INSTEAD OF + Twelve fixes, each on its own + + + + + + A PACKAGE THAT + FAILS ON IMPORT + ▸ DECIDES IT + + + Caught before launch + the run installs the packed release + − + Ships again + 0.1.1 passed every source test + + + + PRICE + − + A heavy closure run + pack, install, serve, real Redis + + + Nothing extra + each child's own test is enough + + + + + A SECURITY FIX + + + Asked for over HTTP + as user B, against A's records + − + Proved one function at a time + never as two users on a server + + + + + + LOCKS IN + The most expensive child in the set + and it repeats until a run files nothing + + + + + FLIPS IF + The release is not going to external users soon + then these are ordinary bugs; the harness waits + + diff --git a/specs/epics/FIX-1635/figures/d2-reproduce-first.svg b/specs/epics/FIX-1635/figures/d2-reproduce-first.svg new file mode 100644 index 0000000000..27f5e87bba --- /dev/null +++ b/specs/epics/FIX-1635/figures/d2-reproduce-first.svg @@ -0,0 +1,100 @@ + + + + + + + D2 · THE CALL + What does a worker do with a ticket written in August? + + + + + CHOSEN + Reproduce first, close with a test + + INSTEAD OF + Implement the fix as written + + + + + + CODE HAS MOVED + SINCE THE TICKET + ▸ DECIDES IT + + + Works on today's code + a closed hole costs one test + − + Rewrites working code + for a hole a later pass closed + + + + PRICE + − + A PR with no fix in it + the test is the whole change + + + One PR, one fix + when the hole is still open + + + + + THE HOLE, AFTER + = + Pinned by a test + red on the commit before it closed + = + Pinned by a test + the fix's own regression test + + + + + + LOCKS IN + A regression test for every hole + closing with a comment alone is not done + + + + + FLIPS IF + The tickets were written against today's main + then reproducing first is a wasted step + + diff --git a/specs/epics/FIX-1635/figures/d3-id-owner-first.svg b/specs/epics/FIX-1635/figures/d3-id-owner-first.svg new file mode 100644 index 0000000000..44a91cc26a --- /dev/null +++ b/specs/epics/FIX-1635/figures/d3-id-owner-first.svg @@ -0,0 +1,112 @@ + + + + + + + D3 · THE CALL + Which principal fix goes first? + + + + + CHOSEN + FIX-1018 binds the id + + INSTEAD OF + FIX-1286 decides first + + INSTEAD OF + Re-key the workspace + + + + + + A CONFIRMED + CROSS-USER READ + ▸ DECIDES IT + + + Closed first + and answers 1286's question + − + Open during review + a design holds the fix + − + Still open + the record is still adoptable + + + + WHO WAITS + − + 1286 and 1634 + implement after 1018 + − + 1018 waits + on 1286's spec + + + Nobody + one key changes + + + + + CALLER IDS + = + Still accepted + scoped to the owner + = + Still accepted + whatever 1286 picks + = + Still accepted + unchanged + + + + + + LOCKS IN + One principal rule, the binding + 1286 may close as proved, with its own test + + + + + FLIPS IF + The shared workspace survives per-owner records + then 1286 decides alone and the edge comes out + + diff --git a/specs/epics/FIX-1635/figures/end-state.svg b/specs/epics/FIX-1635/figures/end-state.svg new file mode 100644 index 0000000000..5eb654cf76 --- /dev/null +++ b/specs/epics/FIX-1635/figures/end-state.svg @@ -0,0 +1,92 @@ + + + + + + + IN THE BOX · WHAT AN APP GETS FROM THE RELEASE + + the release loads + every package imports; + DevTool has its assets + FIX-1431 · FIX-1334 + + an id is an address + another user's request or + session id reaches nothing + FIX-1018 · 1286 · 1022 + + entry points admit + on the stored record and + the target flow's rules + FIX-1021 · 1046 · 1328 + + guards hold + invalid writes throw; + writable: false refuses + FIX-1256 · 1261 · shipped + + queue-host delivery + into an existing session, + under its concurrency + FIX-1634 · Layer 1 + + one turn, one text + non-streaming keeps + every step's text + FIX-1628 + + THE FENCE + no Workforce word in core, engine, client or react · org stays required (FIX-1442) + no named refusal deleted without the guarantee it protected + each hole proved over HTTP, against installed packages, not source + + + + + COMPOSED IN BY THE APP + its principal resolver + who the caller is; the framework + decides what that caller owns + its tenant header + the wall between tenants, + unchanged by this epic + its queue host + BullMQ and Redis, when it + runs work out of process + + + + + REPLACED IN ONE LINE + the resolver: one hook per flow or per host + the re-entry allow-list: publicReentrySources + + + + + NOT BUILT + first-hour DX (FIX-1161) · the Tier B leftovers · Workforce adopting queue delivery · kitchen-sink polish · ops repair + + diff --git a/specs/epics/FIX-1635/figures/ownership.svg b/specs/epics/FIX-1635/figures/ownership.svg new file mode 100644 index 0000000000..b36590a970 --- /dev/null +++ b/specs/epics/FIX-1635/figures/ownership.svg @@ -0,0 +1,125 @@ + + + + WHO OWNS WHAT · ONE OWNER PER RULE + + 1018req id + 1286workspace + 1022session + 1021re-entry + 1046sibling + 1328x-flow + 1431dist + 1334devtool + 1634queue + 1628text + 1636closure + + + + ER-1 · every package imports, packed + builds + uses + + + + ER-2 · DevTool ships its assets + builds + uses + + + + ER-3 · a request id reaches only its owner + builds + uses + uses + uses + + + + ER-4 · no shared run workspace + builds + uses + + + + ER-5 · a session id reaches only its owner + builds + uses + + + + ER-6 · re-entry by allow-list only + builds + uses + + + + ER-7 · a flow never reads a sibling's + builds + uses + + + + ER-8 · cross-flow admits like ingress + builds + uses + + + + ER-9 · an invalid write throws + shipped · FIX-1256 + uses + + + + ER-10 · writable: false refuses + shipped · FIX-1261 + uses + + + + ER-11 · queue host delivers + builds + uses + + + + ER-12 · one turn, one text + builds + uses + + + + ER-18 · FIX-1018 lands first + builds + uses + uses + + + + ER-20 · the proof, on one main commit + builds + + + builds: owns and ships it · uses: consumes it as is, must not re-decide · shipped rows are re-verified by the closure + diff --git a/specs/epics/FIX-1635/figures/path.svg b/specs/epics/FIX-1635/figures/path.svg new file mode 100644 index 0000000000..4fcfe4fca5 --- /dev/null +++ b/specs/epics/FIX-1635/figures/path.svg @@ -0,0 +1,106 @@ + + + + THE PATH · AS OF 29 SEP 2026 · UNSCHEDULED, SO THE AXIS IS ORDER, NOT DATES + + + + shipped + + step 1 + + step 2 + + step 3 + + now + + + + shipped · 1256 · 1261 · 1510 + + merged + re-verified by the closure run, not reopened + + + this epic spec + + review + nothing starts until it merges + + + + + FIX-1018 · request id + + reproduce → fix + + + 1022 · 1021 · 1046 · 1328 + + reproduce → fix or test + + + 1431 · 1334 · the release + + reproduce → fix or test + + + FIX-1628 · text + + fix + + + FIX-1286 · workspace + + spec + + impl + + + FIX-1634 · queue delivery + + spec · Layer 1 + + impl + + + FIX-1636 · closure → wrap + + QA plan + + run → PR → wrap + + + + + + + + + Critical path: FIX-1018 → FIX-1634 impl → FIX-1636 run on one main commit → closure PR merged → wrap. + FIX-1286 and FIX-1634 wait for FIX-1018 only to implement. Green bars are merged; dashed bars are not started. + From 46870d989b594a15583162143cae6725745f4c19 Mon Sep 17 00:00:00 2001 From: Claude Date: Tue, 29 Sep 2026 12:28:08 +0000 Subject: [PATCH 2/3] spec(epic): fold review round 1 into FIX-1635 Leg a becomes a standing CI job (FIX-1431 lands it, FIX-1334 extends it, the closure reuses it) with its control pinned to the 0.1.1 tarball. Each child lands its HTTP case and proves its control at fix time; the closure runs the suite on installed tarballs and no longer rebuilds old commits. ER-3 covers stream attach and resume and asserts owner-scoped reuse; ER-10 matches the shipped create-if-absent contract; the request-as-capability model gets an evolution row owned by FIX-1018. Rules now each carry an owner and a check; ER-18 (a restatement of D3) is dropped and the rest renumbered. DOCS keeps intent and ownership; FIX-1018 owns the draft. Co-Authored-By: Claude Opus 5.5 (1M context) Claude-Session: https://claude.ai/code/session_01HNTQkGXCUryuVD2TEBLi6G --- specs/epics/FIX-1635/BUSINESS-RULES.md | 47 +++++++++++---------- specs/epics/FIX-1635/DECISIONS.md | 27 ++++++++++-- specs/epics/FIX-1635/DOCS.md | 48 +++++++++++----------- specs/epics/FIX-1635/EVOLUTION.md | 1 + specs/epics/FIX-1635/PLAN.md | 13 +++--- specs/epics/FIX-1635/SPEC.md | 26 ++++++------ specs/epics/FIX-1635/figures/ownership.svg | 22 ++++------ 7 files changed, 99 insertions(+), 85 deletions(-) diff --git a/specs/epics/FIX-1635/BUSINESS-RULES.md b/specs/epics/FIX-1635/BUSINESS-RULES.md index 9fa27aaf0b..aea6392f29 100644 --- a/specs/epics/FIX-1635/BUSINESS-RULES.md +++ b/specs/epics/FIX-1635/BUSINESS-RULES.md @@ -9,39 +9,38 @@ run over real HTTP against installed tarballs. | # | Rule | Owner | Checked at | |---|---|---|---| -| ER-1 | Every publishable package, packed and installed into an empty ESM project, imports | FIX-1431 | Closure leg a · CI's extension check | -| ER-2 | DevTool cannot be published without its client assets, and serves them from an installed copy | FIX-1334 | Closure leg a · the release build | -| ER-3 | A request id a caller supplies never reaches another user's request record or its items, on any path that writes one | FIX-1018 | Closure leg b · host and SQLite tests | -| ER-4 | Two users in one tenant with the same request id never share a run-scoped workspace or its resources | FIX-1286 (consumes ER-3) | Closure leg b | -| ER-5 | A session id reaches only its owner's session record; another user's id is not found | FIX-1022 | Closure leg b | -| ER-6 | Only an allow-listed source is re-entered from public HTTP by retry, continue or resume | FIX-1021 | Closure leg b | -| ER-7 | A session's stored flow never authorizes reading another flow's requests or items | FIX-1046 | Closure leg b | -| ER-8 | A cross-flow dispatch admits the target exactly as ingress would, organization included | FIX-1328 | Closure leg b | +| ER-1 | Every publishable package, packed and installed into an empty ESM project, imports | FIX-1431 | Leg a, a standing CI job FIX-1431 lands · the closure reuses it | +| ER-2 | DevTool cannot be published without its client assets, and serves them from an installed copy | FIX-1334 | Leg a, which FIX-1334 extends · the release build | +| ER-3 | A request id a caller supplies never reaches another user's request record or its items, on any path that writes it, attaches to its stream or resumes it. Reusing another user's id gets the caller its own request | FIX-1018 | Its HTTP case · host and SQLite tests · closure leg b | +| ER-4 | Two users in one tenant with the same request id never share a run-scoped workspace or its resources | FIX-1286 (consumes ER-3) | Its HTTP case · closure leg b | +| ER-5 | A session id reaches only its owner's session record; another user's id is not found | FIX-1022 | Its HTTP case · closure leg b | +| ER-6 | Only an allow-listed source is re-entered from public HTTP by retry, continue or resume | FIX-1021 | Its HTTP case · closure leg b | +| ER-7 | A session's stored flow never authorizes reading another flow's requests or items | FIX-1046 | Its HTTP case · closure leg b | +| ER-8 | A cross-flow dispatch admits the target exactly as ingress would, organization included | FIX-1328 | Its HTTP case · closure leg b | | ER-9 | A schema-invalid resource write throws and leaves state as it was | FIX-1256 (shipped) | Closure re-runs its test | -| ER-10 | `writable: false` on a collection refuses every write-shaped path, `create` with `replace` and `delete` included | FIX-1261 with FIX-1510 (shipped) | Closure re-runs its tests | -| ER-11 | On a queue host, a delivery into an existing session runs the recipient under its concurrency policy and incarnation guard; what still can't be honoured is refused by name | FIX-1634 | Closure leg c | -| ER-12 | A non-streaming multi-step turn returns the text the streamed turn would | FIX-1628 | FIX-1628's test · closure re-run | +| ER-10 | `writable: false` on a collection refuses every write to an existing instance, `create` with `replace` included, and `delete`. Creating a key that does not exist stays open | FIX-1261 with FIX-1510 (shipped) | Closure re-runs its tests | +| ER-11 | On a queue host, a delivery into an existing session runs the recipient under its concurrency policy and incarnation guard; what still can't be honoured is refused by name | FIX-1634 | Its HTTP case, real Redis · closure leg c | +| ER-12 | A non-streaming multi-step turn returns the text the streamed turn would | FIX-1628 | Its HTTP case · closure leg b | ## What no child may do -| # | Rule | Because | -|---|---|---| -| ER-13 | No Workforce (Layer 2) concept in core, engine, client or react. FIX-1634 is generic; its Workforce adopt child is filed after its Layer 1 shape merges | The owner's layer rule | -| ER-14 | No child reopens org-optional, or deletes a named refusal without restating the guarantee it protected | FIX-1442 is settled; a deleted refusal trades a loud failure for a silent one | -| ER-15 | No child widens: Tier B leftovers, first-hour DX and kitchen-sink polish stay out. Something noticed outside the goal is filed normally, not under this epic | The set is locked by the owner | -| ER-16 | No child pins today's defect as a contract in a test, and every fix lands a regression test that fails on the commit before the hole closed | A test that can't fail proves nothing | +| # | Rule | Owner · checked at | Because | +|---|---|---|---| +| ER-13 | No Workforce (Layer 2) concept in core, engine, client or react. FIX-1634 is generic; its Workforce adopt child is filed after its Layer 1 shape merges | FIX-1634, the one child near it · its spec review | The owner's layer rule | +| ER-14 | No child reopens org-optional, or deletes a named refusal without restating the guarantee it protected | Every child · FIX-1634's spec names what FIX-1302's `external-dispatcher` refusal guarantees, since it narrows it; each PR review | FIX-1442 is settled; a deleted refusal trades a loud failure for a silent one | +| ER-15 | No child widens: Tier B leftovers, first-hour DX and kitchen-sink polish stay out. Something noticed outside the goal is filed normally, not under this epic | The epic coordinator · each child's PR review against the set table | The set is locked by the owner | +| ER-16 | Every fix lands a regression test that fails on the commit before the hole closed, and no test pins today's defect as a contract. A child named in ER-3 to ER-8, ER-11 or ER-12 lands it as its case in the shared HTTP suite: over HTTP, as the second user where there is one | Each child · its PR names the commit the case failed on | A test that can't fail proves nothing; a control proved at fix time needs no old commit rebuilt at closure | ## How the set is run -| # | Rule | Because | -|---|---|---| -| ER-17 | A direct-route worker reproduces on current `main` first. No reproduction: it closes with a regression test (ER-16) and the commit that closed the hole, not a rewrite | D2 | -| ER-18 | FIX-1018 merges before FIX-1286 and FIX-1634 implement; their specs may start now | D3 | -| ER-19 | A child's Linear state is mirrored the moment it changes; its route comes from its label | The wake derives blocked-by and route from Linear | +| # | Rule | Owner · checked at | Because | +|---|---|---|---| +| ER-17 | A direct-route worker reproduces on current `main` first, and records the result on its Linear issue: reproduces or not, on which commit. No reproduction: it closes with a regression test (ER-16) and the commit that closed the hole, not a rewrite | Each direct-route child · the Linear comment, before its PR opens | D2. The closure cites the record rather than re-deriving it | +| ER-18 | A child's Linear state is mirrored the moment it changes; its route comes from its label | The epic coordinator · every wake | The wake derives blocked-by and route from Linear | ## The closure | # | The epic is done when | Proved by | |---|---|---| -| ER-20 | [The goal](SPEC.md#the-goal-and-how-well-know-its-met) is met: legs a, b and c pass on one `main` commit, each fails under its control, and every finding is fixed and retested | FIX-1636's goal check | -| ER-21 | The authentication guide says a caller-chosen id is an address, not an ownership | FIX-1018's docs change ([DOCS.md](DOCS.md)) | +| ER-19 | [The goal](SPEC.md#the-goal-and-how-well-know-its-met) is met on one `main` commit: leg a passes in its standing job and fails against the 0.1.1 tarball; the HTTP suite (legs b and c) passes against the installed tarballs; every finding is fixed and retested | FIX-1636's goal check | +| ER-20 | The authentication guide says a caller-chosen id is an address, not an ownership, and the architecture's request-id section no longer calls a request id a capability | FIX-1018's docs change ([DOCS.md](DOCS.md)) | diff --git a/specs/epics/FIX-1635/DECISIONS.md b/specs/epics/FIX-1635/DECISIONS.md index 1af2c96212..b30ab74d3a 100644 --- a/specs/epics/FIX-1635/DECISIONS.md +++ b/specs/epics/FIX-1635/DECISIONS.md @@ -3,7 +3,8 @@ [Spec](SPEC.md) · **Decisions** · [Rules](BUSINESS-RULES.md) · [Plan](PLAN.md) · [Docs](DOCS.md) · [Evolution](EVOLUTION.md) The calls that sit above any single child. The child set itself is not one of them: the owner -locked it with the FSD Architect on 2026-09-29, including what is out of the epic. Two cards +locked it with the FSD Architect on 2026-09-29, including what is out of the epic. "Twelve" +below is that locked set of child issues; FIX-1510 shipped as FIX-1261's sub-issue. Two cards are the sign-off; the third is an engineering call recorded so no child reopens it. ## The tree @@ -65,10 +66,12 @@ It comes down to a confirmed leak: the design question should not hold the fix f ## Who owns what -![Who owns what: fourteen rules by the open children and the closure issue. Each security, integrity, packaging, queue and generator rule has one child that builds it; FIX-1286 and FIX-1634 consume FIX-1018's binding; the closure issue builds the proof and consumes every other rule.](figures/ownership.svg) +![Who owns what: thirteen rules by the open children and the closure issue. Each security, integrity, packaging, queue and generator rule has one child that builds it; FIX-1286 and FIX-1634 consume FIX-1018's binding, and FIX-1334 builds on FIX-1431's leg a job; the closure issue runs the children's suite on installed tarballs and consumes every other rule.](figures/ownership.svg) -Every rule has one owner. A *consumes* cell is a place a child must not re-decide: FIX-1286 -and FIX-1634 build on FIX-1018's binding rather than inventing their own. +Every rule a child builds has one owner. A *consumes* cell is a place a child must not +re-decide: FIX-1286 and FIX-1634 build on FIX-1018's binding rather than inventing their own. +The fence and run rules bind every child, so they are not columns here; each names its owner +and where it's checked in [BUSINESS-RULES.md](BUSINESS-RULES.md#what-no-child-may-do). ## Decided in review, recorded so no child reopens them @@ -83,6 +86,18 @@ and FIX-1634 build on FIX-1018's binding rather than inventing their own. - **FIX-1628 stays scoped** to the non-streaming step loop. - **The done rows are re-verified, not reopened.** The closure plan re-runs their tests; a failure is a new bug child. +- **Leg a is a standing CI job, not a closure step** (round 1). FIX-1431 lands the pack, + install and import job and FIX-1334 adds DevTool's assets to it; the closure reuses it. Its + control is the 0.1.1 tarball, pinned, because "the latest release" stops failing once + anything newer is published. +- **Each child owns its HTTP case and its control** (round 1). The children with a hole or a + behaviour to prove over HTTP add a case to one shared suite, and prove it fails on the commit + before the hole closed. The closure runs the suite against installed tarballs and never + rebuilds an old commit. +- **No shared owner-check rule** (round 1). The principal children keep their parallel + bindings and one not-found shape. A shared owner-scoped lookup is FIX-1018's to offer if its + reproduction produces one; the others may use it, and nothing waits for it. +- **FIX-1286's route stays spec** (round 1). D3 already lets it close as proved by FIX-1018. ## What the end-state POC showed @@ -93,5 +108,9 @@ is the whole-set proof. - **Drafted (Sep 29)** from the owner's locked set and the Architect's clusters; the closure issue FIX-1636 filed; FIX-1018 wired to block FIX-1286 and FIX-1634. +- **Round 1 (Sep 29)** folded a second-look review, the Architect's stamp on it, and the Cursor + and Codex inline comments: leg a moved into standing CI, the children own their HTTP cases, + ER-3 names attach and resume, ER-10 matches the shipped create-if-absent contract, and the + rules were trimmed to ones with an owner. **Open: none.** diff --git a/specs/epics/FIX-1635/DOCS.md b/specs/epics/FIX-1635/DOCS.md index bb25f0e842..5a4e734344 100644 --- a/specs/epics/FIX-1635/DOCS.md +++ b/specs/epics/FIX-1635/DOCS.md @@ -6,36 +6,34 @@ One shared story changes: which ids a caller may choose, and what choosing one g Most children change behaviour the docs already promise (a package imports, a guard refuses), so they publish nothing new. The specifics below belong to the child that ships them. -## UPDATE · `apps/docs/docs/server/authentication.md` · new subsection after "Addressed routes, and what they scope by" - -> ### Ids you choose are addresses -> -> A client can pick its own ids. An action call may carry a `requestId`, and most routes name -> a session by id. That is useful: a retried call with the same `requestId` is recognised as -> the same request instead of starting a second one. -> -> An id is never proof of ownership. A request or a session belongs to the user and -> organization it was created under, and an id someone else created reaches nothing of yours. -> Send another user's `requestId` and you get your own request, not theirs. Send another -> user's session id and you get a `404`, the same answer as an id that doesn't exist. -> -> ``` -> POST /api/flows/chat/actions/send bob's, requestId req_1 (alice's) -> bob's own request -> GET /api/flows/sessions/s_alice bob's, same tenant -> 404 -> ``` -> -> So you don't need to make ids unguessable to keep users apart. Ids still leak in headers, -> URLs and logs, and nothing here relies on them staying secret. - -The example routes are illustrative until FIX-1018's `DOCS.md` fixes the exact paths and -status codes against the shipped behaviour. +## The shared promise · UPDATE `apps/docs/docs/server/authentication.md` + +A new subsection after "Addressed routes, and what they scope by", **"Ids you choose are +addresses"**. What it has to tell a reader: + +- A client may choose its own ids: a `requestId` on an action call, a session id on most + routes. A retried call with the same `requestId` is the same request, not a second one. +- An id is never proof of ownership. Another user's `requestId` gets the caller its own + request; another user's session id is a `404`, the same as an id that doesn't exist. +- So ids need not be unguessable to keep users apart. They leak in headers, URLs and logs, + and nothing relies on them staying secret. + +FIX-1018 writes the prose and its example routes, against the behaviour it ships, in its own +PR. This document holds the intent and who publishes what, not a second copy of the draft. + +## UPDATE · `docs/architecture/state-and-scopes.md` · "What `requestId` gates, not tenant" + +The section calls a request id an unguessable capability that authorizes stream attach and +resume alone. After FIX-1018 it is an address scoped to its owner, and the section says what +attach and resume check instead ([EVOLUTION.md](EVOLUTION.md)). ## Ownership | Material | Publisher | Specific draft | |---|---|---| -| "Ids you choose are addresses", above | FIX-1018, once its behaviour is verified | This document | -| The session-id half of the example, and legacy session keys | FIX-1022 | Its PR; links here | +| "Ids you choose are addresses", above | FIX-1018, once its behaviour is verified | FIX-1018's PR | +| The request-id section of `state-and-scopes.md` | FIX-1018 | FIX-1018's PR | +| The session-id half of the example, and legacy session keys | FIX-1022 | Its PR; links to FIX-1018's subsection | | A session's flow does not authorize another flow's history | FIX-1046 | Its PR; one sentence in "Addressed routes" | | The `external-dispatcher` row in `server/background-work.md`'s refusal table, narrowed to what a queue host still can't honour | FIX-1634 | Its `DOCS.md` | | `writable` on collections | Shipped with FIX-1510 | None | diff --git a/specs/epics/FIX-1635/EVOLUTION.md b/specs/epics/FIX-1635/EVOLUTION.md index 6a53f99a66..1ae2cb156b 100644 --- a/specs/epics/FIX-1635/EVOLUTION.md +++ b/specs/epics/FIX-1635/EVOLUTION.md @@ -10,6 +10,7 @@ FIX-1286's own specs. | FIX-1442: every session and request carries an organization, owned by the verified principal. Source: its Linear spec document and review PR #1899 (no retained repository set) | **Retained** | Shipped (Done, 2026-09). FIX-1328's second option is this rule; the principal cluster builds on it | ER-8, ER-14 | Nothing changes for org identity | | FIX-682: session keys and re-entry scoped by tenant. Source: `packages/engine/src/stores/scope-keys.ts` → `resolveSessionStorageKey` | **Amended** by FIX-1022 | The tenant wall holds; users inside one tenant were never separated | ER-5 | Existing tenant-only keys are still read by their owner (BP-030) | | FIX-999: public re-entry admits an allow-list of sources. Source: `packages/engine/src/routes/public-reentry.ts` | **Retained**; FIX-1021 verifies against it | The allow-list landed after FIX-1021 was written | ER-6 | None | +| The request-as-capability model: stream attach and suspension resume are authorized by `requestId` alone, as an unguessable token, with no tenant recheck. Source: `docs/architecture/state-and-scopes.md`, "What `requestId` gates, not tenant" | **Amended** by FIX-1018 | The premise fails: a request id is caller-supplied (`routes/action-routes.ts`), `generateId` is not a CSPRNG, and ids leak in the `x-request-id` header, URLs and logs (FIX-1018's evidence) | ER-3, ER-20; FIX-1018 updates the section | A request created before FIX-1018 still attaches and resumes for its owner (BP-030) | | FIX-1302: delivery into an existing session is refused `external-dispatcher` on a queue host. Source: `docs/architecture/inbound-transports.md`, the `external-dispatcher` paragraph | **Amended** by FIX-1634 | The refusal protected recipient concurrency and the incarnation guard; FIX-1634 restores both across the queue boundary instead of deleting it | ER-11, ER-14 | Anything still unhonourable keeps a named refusal | Nothing is wholly superseded. Re-check each cited intent against current code before diff --git a/specs/epics/FIX-1635/PLAN.md b/specs/epics/FIX-1635/PLAN.md index 18ea9ee791..b6f4002185 100644 --- a/specs/epics/FIX-1635/PLAN.md +++ b/specs/epics/FIX-1635/PLAN.md @@ -18,17 +18,17 @@ The dependency graph is in [the spec](SPEC.md#how-the-issues-flow-into-each-othe | Issue | Route | Consumes | Delivers | Releases | Size | |---|---|---|---|---|---| -| **FIX-1018** request id | direct | Current `main` (ER-17) | ER-3; the docs line (ER-21) | FIX-1286 · FIX-1634 | Medium | +| **FIX-1018** request id | direct | Current `main` (ER-17) | ER-3, and its case in the HTTP suite; the docs line (ER-20) | FIX-1286 · FIX-1634 | Medium | | **FIX-1286** run workspace | spec → impl | ER-3 | ER-4, or a test proving ER-3 already delivers it | The closure | Small, or none | | **FIX-1022** session key | direct | Current `main` | ER-5, with legacy keys still read by their owner | The closure | Medium | | **FIX-1021** re-entry | direct | Current `main` | ER-6 | The closure | Small, likely a test | | **FIX-1046** sibling flow | direct | Current `main` | ER-7 | The closure | Small | | **FIX-1328** cross-flow admission | direct | FIX-1442's org rule | ER-8 | The closure | Small, likely a test | -| **FIX-1431** `dist` extensions | direct | Current `main` | ER-1 | The closure | Small, likely a test | -| **FIX-1334** DevTool assets | direct | Current `main` | ER-2 | The closure | Small, likely a test | +| **FIX-1431** `dist` extensions | direct | Current `main` | ER-1, and leg a as a standing CI job, controlled by the 0.1.1 tarball | FIX-1334's check · the closure | Small, plus the job | +| **FIX-1334** DevTool assets | direct | Leg a's job | ER-2, as a check in leg a | The closure | Small, likely a test | | **FIX-1634** queue delivery | spec → impl | ER-3 · FIX-1302's refusal | ER-11, Layer 1 only | The closure; later, a Workforce adopt child | Large | | **FIX-1628** non-streaming text | direct | FIX-1626's join rule | ER-12 | The closure | Small | -| **FIX-1636** closure · required | spec (the QA plan) → runs until clean → PR | Every other child, merged, on one `main` commit | The committed release check, a QA report, a bug child per finding | Wrap | Large, repeats per retest | +| **FIX-1636** closure · required | spec (the QA plan) → runs until clean → PR | Every other child, merged, on one `main` commit · leg a's job · the children's HTTP suite | The suite run against installed tarballs with a real Redis, a QA report, a bug child per finding | Wrap | Large, repeats per retest | "Likely a test" is not a diagnosis. Each worker's **first step is reproducing the ticket's hole on current `main`** (ER-17): FIX-1442, FIX-999's re-entry allow-list, the ESM-extension build @@ -60,7 +60,8 @@ FIX-1628 matches). | Session and request routes' owner checks | FIX-1022, FIX-1046, FIX-1021 | Not-found for another user's record, one shape across all three | | The run-scope key | FIX-1018 and FIX-1286 | 1286 changes it only if 1018's binding leaves a shared workspace | | `server/authentication.md` | FIX-1018, FIX-1022, FIX-1046 | 1018 publishes the id rule; the others link to it | -| The publish path (`release:build`, CI's `dist` check) | FIX-1431 and FIX-1334 | One release build; neither adds a second | +| The publish path (`release:build`, CI's `dist` check) and leg a's job | FIX-1431 and FIX-1334 | One release build and one leg a job: 1431 creates it, 1334 adds to it, neither adds a second | +| The shared HTTP suite | FIX-1018, 1286, 1022, 1021, 1046, 1328, 1634, 1628 and FIX-1636 | One location. The first child to land creates it; each adds only its own case; the closure runs it and never rewrites a case | ## Not children, deliberately @@ -70,6 +71,6 @@ its Layer 1 shape merges. ## Wrap -When ER-20 holds: the lessons pass, a docs polish over the authentication and background-work +When ER-19 and ER-20 hold: the lessons pass, a docs polish over the authentication and background-work pages, and a completion report in Linear. The original merged spec PR stays as the review record. diff --git a/specs/epics/FIX-1635/SPEC.md b/specs/epics/FIX-1635/SPEC.md index e080dd8fd5..45b13e8d4a 100644 --- a/specs/epics/FIX-1635/SPEC.md +++ b/specs/epics/FIX-1635/SPEC.md @@ -2,7 +2,7 @@ **Spec** · [Decisions](DECISIONS.md) · [Rules](BUSINESS-RULES.md) · [Plan](PLAN.md) · [Docs](DOCS.md) · [Evolution](EVOLUTION.md) -Epic · 12 children and a closure issue · Public Launch · Goal 4, keep the foundation honest, +Epic · 12 child issues (FIX-1510 ships under FIX-1261) and a closure issue · Public Launch · Goal 4, keep the foundation honest, with Goal 3 ([`docs/objectives.md`](../../../docs/objectives.md)) · [FIX-1635](https://linear.app/fixpoint-labs/issue/FIX-1635) · child set locked by the owner with the FSD Architect, 2026-09-29 @@ -35,27 +35,29 @@ and none of the holes this epic names can be reached over real HTTP.** ```mermaid flowchart LR - A["pack every package · install into an empty project"] --> L1["leg a · import each, start a server"] - L1 --> L2["leg b · two users, one tenant, one leg per hole over HTTP"] + A["pack every package · install into an empty project"] --> L1["leg a · import each, start a server · standing CI job"] + L1 --> L2["leg b · the children's HTTP suite · two users, one tenant, one case per hole"] L1 --> L3["leg c · BullMQ host with real Redis"] L2 --> R["what each caller gets back"] L3 --> R R -->|"every leg holds"| P["PASS · the epic's goal is met"] - C["control · last npm release, or the commit before each fix"] -.-> R + C["control · the 0.1.1 tarball for leg a"] -.-> R R -.->|"under the control"| F["must FAIL · names its leg"] ``` -Leg a is the one no child's own test reaches. Leg b is where each hole is asked for as an -attacker would ask for it. +Leg a is the one no child's own test reaches, so it becomes a standing CI job the moment +FIX-1431 lands it, not a step that runs once at the end. Leg b is where each hole is asked for +as an attacker would ask for it: each child adds its own case, and the closure runs them all +against the installed tarballs. | How we verify | | |---|---| -| **Goal check** | [FIX-1636](https://linear.app/fixpoint-labs/issue/FIX-1636)'s QA plan, on one `main` commit after every other child merges ([ER-20](BUSINESS-RULES.md#the-closure)) | -| **Model** | Scripted and keyless. The generator leg asserts on text a scripted model wrote, not on a model's quality | -| **Signal** | **a:** every packed package imports in an empty ESM project, and DevTool serves its client assets from the installed copy. **b:** as user B in user A's tenant, each named hole returns not-found or refused, and A's records are unchanged afterwards. **c:** a delivery into an existing session runs the recipient on a queue-backed host | +| **Goal check** | [FIX-1636](https://linear.app/fixpoint-labs/issue/FIX-1636)'s QA plan, on one `main` commit after every other child merges. It reuses the leg a job and runs the children's suite; it builds no leg of its own ([ER-19](BUSINESS-RULES.md#the-closure)) | +| **Model** | Scripted and keyless. FIX-1628's case in leg b asserts on text a scripted model wrote, not on a model's quality | +| **Signal** | **a:** every packed package imports in an empty ESM project, and DevTool serves its client assets from the installed copy. **b:** as user B in user A's tenant, each named hole returns not-found or refused, except a reused request id, which succeeds as B's own request; A's records and items are unchanged afterwards either way. A multi-step non-streamed turn returns the text the streamed turn does. **c:** a delivery into an existing session runs the recipient on a queue-backed host | | **Input** | Ids B learns the way an attacker would: A's request id from a response header, A's session id from a URL | | **Anti-game** | No leg imports from `src`, calls a store or engine function directly, or runs against a workspace link. Each passes while a consumer still breaks | -| **Control that must fail** | Leg a against the last published npm version. Each leg b and c case against the commit just before its fix merged, including fixes that shipped before this epic ([D2](DECISIONS.md#d2)) | +| **Control that must fail** | Leg a against the 0.1.1 tarball, pinned in the job, never "the latest release". Each leg b and c case fails on the commit before its hole closed, including holes closed before this epic ([D2](DECISIONS.md#d2)). The child proves that when it lands the case and records the commit ([ER-16](BUSINESS-RULES.md#what-no-child-may-do)); the closure never rebuilds an old commit | ## What's in the box @@ -72,7 +74,7 @@ spec PR by design. | Issue | What it delivers | Why the set needs it | Route · status | |---|---|---|---| | [FIX-1256](https://linear.app/fixpoint-labs/issue/FIX-1256) · invalid resource write | Throws instead of resetting state | Silent data loss | direct · **Done** · [#1469](https://github.com/fixpoint-labs/flow-state-dev/pull/1469) | -| [FIX-1261](https://linear.app/fixpoint-labs/issue/FIX-1261), remainder [FIX-1510](https://linear.app/fixpoint-labs/issue/FIX-1510) · `writable: false` on collections | Refused on every write-shaped path | A guard a neighbour bypasses | direct · **Done** · [#1470](https://github.com/fixpoint-labs/flow-state-dev/pull/1470), [#2047](https://github.com/fixpoint-labs/flow-state-dev/pull/2047) | +| [FIX-1261](https://linear.app/fixpoint-labs/issue/FIX-1261), remainder [FIX-1510](https://linear.app/fixpoint-labs/issue/FIX-1510) (its sub-issue) · `writable: false` on collections | Refused on every write to an existing instance, and on delete; creating a missing key stays open | A guard a neighbour bypasses | direct · **Done** · [#1470](https://github.com/fixpoint-labs/flow-state-dev/pull/1470), [#2047](https://github.com/fixpoint-labs/flow-state-dev/pull/2047) | | [FIX-1018](https://linear.app/fixpoint-labs/issue/FIX-1018) · request id reparenting | A caller's request id cannot reach another user's record or items | Confirmed cross-user read; first of the principal cluster ([D3](DECISIONS.md#d3)) | direct · Backlog | | [FIX-1286](https://linear.app/fixpoint-labs/issue/FIX-1286) · run-scoped workspace | Two users with one request id never share a sandbox | Its issue asks for a decision, not a patch | spec · Backlog · blocked by 1018 | | [FIX-1022](https://linear.app/fixpoint-labs/issue/FIX-1022) · session key has no principal | Another user's session id is not found | Same-tenant session read | direct · Todo | @@ -85,7 +87,7 @@ spec PR by design. | [FIX-1628](https://linear.app/fixpoint-labs/issue/FIX-1628) · non-streaming text drop | A multi-step turn keeps every step's text | Same turn, two answers | direct · Backlog | | [FIX-1636](https://linear.app/fixpoint-labs/issue/FIX-1636) · closure · **required** | The QA plan, run on one `main` commit | The only child that runs the release as a consumer does | spec · Backlog · blocked by every open child | -**3 done · 10 open · the closure.** Could it be smaller? The set is the owner's, locked with +**12 children: 2 done, 10 open, plus FIX-1510 done under FIX-1261 · the closure.** Could it be smaller? The set is the owner's, locked with the Architect, and this spec does not reopen it. What can shrink is the work: several tickets predate later passes on `main`, so each direct-route worker reproduces first, and a hole that is already closed closes with a test instead of a rewrite ([D2](DECISIONS.md#d2)). diff --git a/specs/epics/FIX-1635/figures/ownership.svg b/specs/epics/FIX-1635/figures/ownership.svg index b36590a970..641519c82c 100644 --- a/specs/epics/FIX-1635/figures/ownership.svg +++ b/specs/epics/FIX-1635/figures/ownership.svg @@ -1,5 +1,5 @@ - + - + WHO OWNS WHAT · ONE OWNER PER RULE 1018req id @@ -38,6 +38,7 @@ ER-1 · every package imports, packed builds + uses uses @@ -109,17 +110,10 @@ uses - - ER-18 · FIX-1018 lands first - builds - uses - uses - - - ER-20 · the proof, on one main commit - builds - + ER-19 · the suite, on installed tarballs + builds + - builds: owns and ships it · uses: consumes it as is, must not re-decide · shipped rows are re-verified by the closure + builds: owns and ships it · uses: consumes it as is, must not re-decide · shipped rows are re-verified by the closure From 510e2411ab81c11507a3fdee1caf8933449fe218 Mon Sep 17 00:00:00 2001 From: Claude Date: Tue, 29 Sep 2026 12:29:07 +0000 Subject: [PATCH 3/3] spec(epic): FIX-1635 plan names where the closure's Redis comes from Co-Authored-By: Claude Opus 5.5 (1M context) Claude-Session: https://claude.ai/code/session_01HNTQkGXCUryuVD2TEBLi6G --- specs/epics/FIX-1635/PLAN.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/specs/epics/FIX-1635/PLAN.md b/specs/epics/FIX-1635/PLAN.md index b6f4002185..fa78894450 100644 --- a/specs/epics/FIX-1635/PLAN.md +++ b/specs/epics/FIX-1635/PLAN.md @@ -28,7 +28,7 @@ The dependency graph is in [the spec](SPEC.md#how-the-issues-flow-into-each-othe | **FIX-1334** DevTool assets | direct | Leg a's job | ER-2, as a check in leg a | The closure | Small, likely a test | | **FIX-1634** queue delivery | spec → impl | ER-3 · FIX-1302's refusal | ER-11, Layer 1 only | The closure; later, a Workforce adopt child | Large | | **FIX-1628** non-streaming text | direct | FIX-1626's join rule | ER-12 | The closure | Small | -| **FIX-1636** closure · required | spec (the QA plan) → runs until clean → PR | Every other child, merged, on one `main` commit · leg a's job · the children's HTTP suite | The suite run against installed tarballs with a real Redis, a QA report, a bug child per finding | Wrap | Large, repeats per retest | +| **FIX-1636** closure · required | spec (the QA plan) → runs until clean → PR | Every other child, merged, on one `main` commit · leg a's job · the children's HTTP suite | The suite run against installed tarballs with a real Redis (its QA plan says where CI gets it), a QA report, a bug child per finding | Wrap | Large, repeats per retest | "Likely a test" is not a diagnosis. Each worker's **first step is reproducing the ticket's hole on current `main`** (ER-17): FIX-1442, FIX-999's re-entry allow-list, the ESM-extension build