Repository navigation
spec(epic): the public release installs, and holds its boundaries (FIX-1635) - #2360
Conversation
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) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HNTQkGXCUryuVD2TEBLi6G
|
The latest updates on your projects. Learn more about Vercel for GitHub. 3 Skipped Deployments
|
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
Second-look + architecture review (spec PR #2360)Scope note: this is an epic-spec, so this review stays at set level. Feedback that touches one child is marked → implementer note and belongs on that child's spec/PR, not here. Nothing below blocks the direction; items 1–3 are the ones I'd fold into the spec. Second look: is there a leaner or better shape?1. Leg a should be a standing CI gate, not just a closure step (fold in). 2. Leg a's control ("last published npm version") is unstable. 3. Leg b's controls and ER-16 describe two overlapping tests per hole (fold in). 4. FIX-1286 as spec→impl may be paying for a design PR the answer might not need. (→ decision for the epic wake, low stakes) Architecture: is this an opportunity to consolidate rather than add three copies?5. Owner-scoped lookup should be one primitive owned by FIX-1018 (fold in). 6. Same for ER-8 (cross-flow admission "exactly as ingress would"). (→ implementer note for FIX-1328) Smaller notes (non-blocking)
Net: direction is sound and the D1 argument (0.1.1 passed every source test) is convincing. The main improvement is moving the cheap, durable check (leg a) earlier and standing, letting children own their HTTP tests, and consolidating the owner-check into one primitive instead of three parallel bindings. Generated by Claude Code |
jhoffner
left a comment
There was a problem hiding this comment.
Coherence check (FSD Architect) — not owner approval / not the gate. Shared jhoffner login can look like product-owner sign-off; this is approach/coherence only. Spec/epic gate already passed; this stamp is holistic fork-check only.
Verdict: No two-direction fork. This is the sole open spec(epic) for FIX-1635 / hard gates. It sits cleanly under #2359 (Public Launch project-spec, stamped coherent earlier today) and does not compete with it.
Matches locked graph:
- Child set — Jake+Architect lock (1634, 1431, 1018, 1286, 1021, 1022, 1256, 1510+1261 as one outcome, 1046, 1328, 1334, 1628) plus proof-child FIX-1636. Linear nesting agrees; 1510 stays under 1261, not a 13th feature.
- Placement — Public Launch cross-cut; invent-kills nest under FIX-1161 / Done FIX-1592; soft-related FIX-1161 only as parallel first-hour track.
- Closure framing — FIX-1636 = pack→install→HTTP proof-child / mayWrap (1596/1601 spine), not a feature gate. D1/D2/D3 and the stranger-harm fence match PD-2 on #2359.
Does not fork: no second hard-gates epic-spec; #1312 (FIX-548) stays first-hour; Tier B / KS polish / Ops / L2 adopt for 1634 correctly fenced out. Soft note only: #2359’s “2 of 12 children done” snapshot lags this PR’s accurate “3 done” (1256/1261/1510) — refresh on the project-spec when convenient; not a product conflict.
No Jake ping. Coherence/approach only — human owner still owns merge of this never-merge epic-spec.
There was a problem hiding this comment.
Simplify review (spec-only PR)
Verdict: approve direction. Scope, D2 (reproduce-first), and D3 (FIX-1018 before 1286/1634) are the right complexity reducers. The locked issue set and HTTP consumer bar match RELEASING.md and the 0.1.1 lesson. I would not reopen the twelve-issue bundle.
The main place the approach risks more machinery than the fixes is not the epic framing—it is how FIX-1636 is allowed to land. D1 explicitly accepts a large, end-loaded closure child; that is defensible for a one-time pre-release gate, but the repo already has most of the ingredients to avoid a monolithic “run everything once at the end” harness.
What to keep
- D2 / ER-17 — best part of the set; shrinks work without shrinking guarantees.
- Leg b anti-game (HTTP only, two users, one tenant) — correct bar; implement via
goals/+ router patterns (hire-plane,cli-principal.integration.test.ts), not new ad-hoc harness code. - FIX-1601 +
orchestration.mdclosure template — reuse structure for FIX-1636’s plan PR; do not invent a second closure doctrine.
Simpler architecture (optional epic amendment; otherwise FIX-1636 child spec)
| Current shape | Lighter alternative |
|---|---|
| FIX-1636 blocked until all children merge; full plan rerun per finding batch | Incremental main CI jobs added as children land: (1) release:build → pack publish set → empty ESM install/import + DevTool asset probe (leg a); (2) matrix HTTP security cases sharing one server (leg b); (3) Redis service job for BullMQ (leg c). Epic wrap = those checks green + one SHA, not a separate QA-only integrator. |
| ER-1–12 + closure re-run + child tests | One “checked at” column: required CI job or package test; avoid “child test and closure leg” duplication unless leg b truly walks a different path. |
| Controls: pre-fix commit per hole + full rebuild each time | Build artifact once per baseline SHA (worktree/matrix assertions only); do not nest release:build + ~27 packs inside every control case. |
| 21 ERs (many restate DECISIONS / SPEC fence) | Keep ER-1–12 as the traceability table; fold ER-13–19 process rules into PLAN or drop ER-19 (orchestration); ER-20/21 duplicate SPEC/DOCS. Nit—follow-up OK. |
| Six decision/path SVGs + parallel mermaid | One source of truth for graphs; SVGs drift from the dated issue table. Nit. |
Third-party / existing tooling
- Do not duplicate
add-esm-extensions.mjs --check --all,publint, or dry-run publish as “closure”—they already pass on broken 0.1.1. - Do script the install loop already documented in
RELEASING.md(possibly onegoals/release-gates/goal), committed on the closure PR per orchestration. packages/integration-testsstays wrong for leg b (no HTTP; by design)—don’t force scenarios there.
Skipped (owner call, not simplify nits)
- Shrinking the locked twelve-issue set.
- Whether leg c is release-critical vs post-1634 canary (spec treats it as in-box).
What I did not flag
No code to run tests on. Spec prose and figure count are in line with sibling epics (e.g. FIX-1592/1457).
Merge recommendation: merge this epic spec; when FIX-1636’s QA plan is written, bias toward early CI wiring + FIX-1601-shaped goals rather than a finale-only manual closure loop. If you want D1 amended to say “closure checks live on main before wrap,” that would be a small DECISIONS/SPEC tweak—not required for me to approve the direction.
Sent by Cursor Automation: Review For Best Approach
|
Coherence check (FSD Architect) — not owner approval / not the gate. Shared Verdict: Direction coherent with Public Launch and the locked child set. No silent duplicate epic and no two-direction fork with another open stream. The outside second-look (Claude Code / Jake login) is constructive set-level feedback; fold-ins 1–3 and 5 need an explicit Architect accept/reject so the PR agent does not invent a second sequencing / ownership story while rewriting. vs other open streams
Claude second-look items (set-level)
Rewrite ask (for the PR agent)Fold 1–3 into SPEC / PLAN / BUSINESS-RULES (and the ownership figure if ER-1 / ER-2 delivery moves to FIX-1431 / FIX-1334 with closure reuse). Do not fold 5 into a new ER owned by FIX-1018, or into new blocked-by edges for FIX-1022 / 1046 / 1021. Leave 4 for the epic wake. Item 6 stays off this epic. No Jake ping — review was already on the PR; this only stamps fold-ins so the rewrite has one direction. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 5132f36601
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
… bot review Status: hard gates' objective is in review (PR #2360, closure FIX-1636 filed); the epics table carries epic state only, counts live once in Now. Folded from #2359 review: check 5 means the showcase paths run, not polish; the docs, brand and demo bar gates the cut on checks 3 and 5 only; framework simplification is the one exception to whatever-landed; a decided-once line for tier 2/3 testing and FIX-107; PR-1's checked where points at FIX-1636's closure run; the territory figure holds no live status; the open 1.0-vs-0.x fork gets its trade-off figure. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HNTQkGXCUryuVD2TEBLi6G
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) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HNTQkGXCUryuVD2TEBLi6G
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HNTQkGXCUryuVD2TEBLi6G
|
Round 1 folded (head 510e241). This covers the second-look review, the Architect's stamp on it, and the Cursor and Codex inline comments. D1, D2 and D3 are unchanged.
🤖 Generated with Claude Code Generated by Claude Code |
jhoffner
left a comment
There was a problem hiding this comment.
Coherence check (FSD Architect) — not owner approval / not the gate. Shared jhoffner login can look like product-owner sign-off; this is approach/coherence only.
Round 1 fold verified on head 510e241 against the Architect rewrite ask (issuecomment-5890190886). Matches. No silent fork and no second sequencing / ownership story.
| Asked | On head |
|---|---|
Fold 1–3 (leg a standing CI via FIX-1431 / FIX-1334; pin control to 0.1.1; children own HTTP suite / ER-16; closure reuses, no old-commit rebuild) |
Present in SPEC goal check, PLAN (FIX-1431 / 1334 / 1636 rows + shared-suite seam), BUSINESS-RULES ER-1/2/16/19, and DECISIONS "Decided in review" |
| Do not fold 5 (no new owner-check ER; no new blocked-by for FIX-1022 / 1046 / 1021) | Held: PLAN still has parallel bindings + one not-found shape; DECISIONS explicitly records "No shared owner-check rule"; optional shared lookup stays FIX-1018 → implementer note |
| Park 4 (FIX-1286 route) | Held: still spec → impl; D3 "may close as proved" unchanged |
| 6 off this epic | Held: implementer note only |
| D1 / D2 / D3 unchanged | Held |
Child count and fence still match the locked set (12 + closure; FIX-1510 under 1261; ER-13/15). Soft only: when #2359 is next edited, its done-count may want a refresh — not a blocker here.
No Jake ping.


Goal: 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.
Leg a is the one no child's own test reaches: 0.1.1 passed every source test. It becomes a standing CI job when FIX-1431 lands it, and each child proves its own HTTP case fails on the commit before its fix. Proved by the closure issue FIX-1636's goal check on one
maincommit. Why this goal: SPEC.md.Inside the box is what any app gets from the release without writing a line. The fence keeps a hard-gates epic from turning into a redesign.
Why now. The release is the first time someone outside this repository runs FSD. 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 set: twelve child issues, locked by the owner with the FSD Architect, plus the closure issue. Request and session ids, re-entry, sibling-flow reads, cross-flow admission, the packed release, queue-host delivery, non-streaming text. As of 2026-09-29: 2 done (FIX-1256, 1261, with FIX-1510 under 1261), 10 open, the closure filed; the set table links live Linear state.
Almost everything starts at the gate. The critical path is FIX-1018, then FIX-1634's implementation, then the closure run.
Every rule a child builds has one owner. FIX-1286 and FIX-1634 build on FIX-1018's binding instead of inventing their own.
Sign off
These twelve are one epic with one release-shaped proof · D1
It comes down to 0.1.1: every source test passed and the first import failed. If wrong: we spend a closure run on fixes that were each fine alone.
FIX-1018 lands first; a caller's request id is bound to its owner · D3
It comes down to a confirmed leak: a design question should not hold its fix. If wrong: FIX-1286 and FIX-1634 wait a few days on a fix they didn't need.
What would change my mind on D1: the release not going to external users soon. Recorded, not asked: D2, each worker reproduces on current
mainfirst, records the result on its Linear issue, and a hole already closed closes with a regression test. Reasoning: DECISIONS.md. Rules: BUSINESS-RULES.md.Open: none. Merging this PR is the approval.
Reviewers · look here
state-and-scopes.md(EVOLUTION.md).main. FIX-1431, 1334, 1021 and 1328 predate passes that look like they fixed them. D2 makes each worker prove it with a test instead of rewriting code.Not here: any single issue's approach. Those are the spec and implementation PRs.
Spec · Decisions · Rules · Plan · Docs · Evolution
· Linear FIX-1635 · Project: Public Launch · merges after human approval and required checks
How to review this — an epic-spec, not an implementation plan
This is an epic-spec: the shared objective and cross-cutting decisions for a set of
issues, in a retained document set. It is not any one issue's implementation design.
In scope to challenge:
SPEC.md— is this body of work worth doing, and is thegoal the real need at full size, or a smaller one the set could meet while the need stays
unmet? Would the closure issue's check fail if the goal were not met?
That question can only be asked here.
DECISIONS.md— shared surface, naming, sequencing, contracts.BUSINESS-RULES.mdwith no owner, or two.DOCS.md; missing or misleadingpredecessor treatment in conditional
EVOLUTION.md.Out of scope — owned by the individual issue specs:
poc/show shape and scoping, not shippable implementation. Directional evidence, experimental
isolation and absence of secrets/generated dependencies remain in scope.
Feedback in the second list is routed to the issue it concerns as an implementer note, not
folded in here.
Required repository checks, approvals and review-thread policy still govern merge.
Optional comments are triaged, not a reason to grind to zero or restart the design review.
🤖 Generated with Claude Code
https://claude.ai/code/session_01HNTQkGXCUryuVD2TEBLi6G