FE-800: Spec to orchestrator plan emitter — generate plan.yaml from a completed specification by kostandinang · Pull Request #167 · hashintel/brunch

kostandinang · 2026-06-03T15:34:17Z

Summary

This PR closes the loop between a confirmed brunch specification and the orchestrator: brunch plan <specId> emits a runnable plan.yaml from a completed spec, multiple specs can coexist on the same project under .brunch/cook/specs/<specId>/plan.yaml, and brunch cook picks which one to run via --spec=<id> or auto-pick.

The emitter is a three-stage projection — deterministic graph read → single LLM planning pass → deterministic reconciliation — and every meaningful transformation surfaces as a typed warning on a single audit stream. Spec-scoped storage is owned by one helper that both the writer and the resolver delegate to.

What Changed

Adds brunch plan <specId> (server-side), driven by a snapshot built from the completed spec's accepted requirements, criteria, and active-path edges.
Adds projection → planning → reconciliation as three pure stages with a typed EmitterWarning audit stream; failure paths fall back to a usable orderless plan.
Emits plans to .brunch/cook/specs/<specId>/plan.yaml so multiple specs coexist on the same project.
Adds brunch cook --spec=<id> with resolution precedence: <dir>/plan.yaml → explicit spec → newest spec by mtime → legacy .brunch/cook/plan.yaml.
Extracts spec-plan-paths as the single owner of the spec-scoped layout, latest-by-mtime selection, and spec-id parsing.
Adds regression oracles for projection determinism, reconciliation, emitter fallback, snapshot building, parser surface, and resolver precedence.

Scope

The generated plan is a reviewable artifact, not a silent input — it round-trips through loadPlan and drives brunch cook --petrinaut-stream end-to-end. The orchestrator package stays pure of server-side code; the server depends on the orchestrator, never the reverse.

Verification

npm run verify passes: format and lint checks, full test suite, production build. Live smoke against the working project confirms brunch plan 23 emits the expected spec-scoped plan and brunch cook --spec=23 resolves it.

Records two 2026-06-03 spikes proving a cook plan.yaml can be projected + planned from a completed intent graph (projection deterministic, ordering via LLM planning pass + reconciliation). Adds PLAN frontier spec-to-cook-plan, SPEC A97 + D160-K. Co-authored-by: Claude <noreply@anthropic.com>

Spike verdict (execution order isn't spec truth; FE-700 won't supply it) weakens Phase 3's premise. Flag FE-800 partially subsumes petri-graph-compilation in both frontier defs + the TRACK F tree; Phase 3 residual value is the Phase-4 simulation oracle. Co-authored-by: Claude <noreply@anthropic.com>

….yaml skeleton Pure projectCookPlanFromSpec(snapshot) → Plan in src/orchestrator/src/cook-plan-projection.ts maps each `requirement` knowledge item to one cook slice with stable `req-<kindOrdinal>` id, populates `verification` from incoming `criterion --verifies--> requirement` edges, attaches every slice to a single default epic, and leaves `depends_on` empty (graph-read execution ordering is intentionally dropped — slice 2's LLM planning pass owns it). 7 acceptance tests: - empty snapshot - N-requirement slice generation (kindOrdinal-ordered, stable ids) - verifies-edge linkage (criteria sorted by kindOrdinal) - depends_on edge between requirements does NOT populate slice.depends_on (regression pin for the deliberate non-goal) - determinism - loadPlan YAML round-trip + schema-conformance (slice.epic_id resolves) - brunch_graphs corpus fixture pinning the spike's "every requirement has ≥1 verifying criterion" oracle as a regression check Local CompletedSpecSnapshot type keeps the orchestrator package independent of @/server/* — the server-side snapshot builder + the orchestrator↔server transport are a separate slice. PLAN.md frontier status → active. CARDS.md slice 1 → done. Amp-Thread-ID: https://ampcode.com/threads/T-019e8dea-2ba3-776f-8d0d-57c94acf5f93 Co-authored-by: Amp <amp@ampcode.com>

…non-buildable detection Pure planExecutionOrdering(plan, runModel) → Promise<PlanningResult> in src/orchestrator/src/cook-plan-llm-planning.ts: - Builds a prompt from the slice-1 projected Plan (instructions for dependsOn DAG inference, epic grouping ~2-5 epics, conservative non-buildable-constraint flagging; lists every available slice id). - Calls injected runModel: (prompt: string) => Promise<unknown> seam. - Parses output through Zod planningEnrichmentSchema (SHAPE only; id-existence, cycles, dangling deps onto constraints deferred to slice 3 reconciliation). - Empty plan short-circuits without an LLM call. - LLM exceptions + parse errors + schema violations all collapse to { status: 'failed', reason: string } so slice 3 can fall back deterministically rather than crashing the pipeline. defaultRunModel uses @ai-sdk/anthropic generateText + Output.object, mirroring src/server/reconciliation-agent.ts:114. Model knob via SPEC_TO_COOK_PLAN_MODEL env, defaulting to claude-sonnet-4-20250514 (matches the 2026-06-03 spike). 7 unit tests with stubbed runModel: - success with well-formed enrichment - failed on thrown runModel - failed on missing required field - failed on wrong-typed field - succeeded on semantically wrong (hallucinated id) — regression pin that slice 2 does NOT enforce existence - prompt content includes every slice id and definition - empty plan short-circuits without invoking runModel 1 opt-in real-LLM integration test gated on PLANNING_REAL_LLM=1 + ANTHROPIC_API_KEY: feeds the brunch_graphs corpus fixture through projection → planning, asserts succeeded + non-trivial signal (at least one ordering edge OR one non-buildable flag). PLAN.md frontier status: slice 3 (deterministic reconciliation) next. CARDS.md slice 2 → done. Amp-Thread-ID: https://ampcode.com/threads/T-019e8dea-2ba3-776f-8d0d-57c94acf5f93 Co-authored-by: Amp <amp@ampcode.com>

…nrichment → cook-runnable Plan Pure `reconcileCookPlan(projected, enrichment)` that turns slice 1's projected Plan plus slice 2's PlanningEnrichment into a cook-runnable Plan + typed warnings: - drop dependsOn refs to nonexistent slice ids - drop self-loops - drop non-buildable slices (preserving definition in the warning) - drop dependsOn edges onto non-buildable slices - cycle-break via Kahn's algorithm with lex-smallest tie-break - assign orphan slices to a synthesized default epic - drop empty LLM-proposed epics - synthesize one `unit-test` verification per surviving slice at `tests/<sliceId>.test.ts`, matching cook's net-compiler reader (`net-compiler.ts:313`) - enrich `slice.definition` with the projected criterion text so the pi-agent has the test context when authoring the test file Every transformation surfaces as a typed `ReconciliationWarning` so a reviewer can audit slice 2's output before it drives cook. 12 unit tests cover each rule plus a brunch_graphs corpus end-to-end that round-trips the reconciled plan through `loadPlan` and a determinism pin. `npm run verify` green. Amp-Thread-ID: https://ampcode.com/threads/T-019e8dea-2ba3-776f-8d0d-57c94acf5f93 Co-authored-by: Amp <amp@ampcode.com>

…, surfaces warnings Composes slices 1, 2, 3 into one end-to-end emitter and exposes it via a new `brunch plan` command. - New `src/orchestrator/src/cook-plan-emitter.ts` exports the pure composition `emitCookPlanFromSnapshot(snapshot, { runModel? })`. On LLM failure the planning result is preserved as `{ status: 'failed', reason }` and reconciliation runs against an empty enrichment so a usable (orderless) plan still emits. - New `src/orchestrator/src/plan-cli.ts` exports `parsePlanArgs` and `runPlan`. Reads a `CompletedSpecSnapshot` JSON, calls the emitter, writes `<outDir>/.brunch/cook/plan.yaml` (creating the directory if missing), and prints every reconciliation warning on stderr with a ` ! ` prefix and a human-readable per-code format. - `src/server/cli.ts` dispatches `brunch plan <snapshot.json> [--out=<dir>] [--verbose]` to `runPlan` and lists it in --help. 9 unit tests cover the composition success path, LLM-failure fallback, YAML round-trip, arg parsing (snapshot path, --out, --verbose / -v, missing-snapshot usage error), end-to-end YAML emission, and warning surfacing. `npm run verify` green (one rerun for the known unrelated `src/server/app.test.ts` flake, as in slices 1/2). Amp-Thread-ID: https://ampcode.com/threads/T-019e8dea-2ba3-776f-8d0d-57c94acf5f93 Co-authored-by: Amp <amp@ampcode.com>

…sis demoted, formatter co-located Addresses ln-review findings #2 (synthesized-target noise), #3 (planning failure across two return shapes), #5 (formatter colocation). - New `reconciliationWarningCategory(w): 'transformation' | 'synthesis'` and `formatReconciliationWarning(w): string` exported from `cook-plan-reconciliation.ts`, both exhaustive over the warning union (build-break enforcement on new codes). - New `EmitterWarning = ReconciliationWarning | { code: 'planning-failed'; reason: string }` widens the audit stream so callers iterate one source instead of forking on `planningResult.status`. The emitter pushes one `planning-failed` warning when the LLM throws and still falls back to reconciliation against an empty enrichment so a usable orderless plan emits. `planningResult` is preserved unchanged for callers that want the raw stage status. - `emitterWarningCategory` adds `'failure'` to the category vocabulary; `formatEmitterWarning` delegates to the reconciliation formatter for non-failure codes. - `runPlan` partitions display by category: failure + transformation always shown under the ` ! ` prefix; synthesis only with `--verbose`. The "N warnings" header counts only what gets printed so screen output stays self-consistent. Smoke-tested against `brunch_graphs` fixture: clean case now emits zero warning lines (was 5 synthesis-noise lines pre-slice). `--verbose` restores the synthesis lines for reviewers who want the full trace. 16 new reconciliation tests (8 category + 8 formatter), 3 new emitter tests (planning-failed presence/absence + category dispatch), 1 new + 2 reshaped plan-cli tests (verbose toggle + planning-failed in `!`-stream). `npm run verify` green on first try. Amp-Thread-ID: https://ampcode.com/threads/T-019e8dea-2ba3-776f-8d0d-57c94acf5f93 Co-authored-by: Amp <amp@ampcode.com>

…ommand) The plan emitter is an orchestrator-package capability that happens to produce a YAML the cook command later consumes. Prefixing the modules with `cook-` framed the artifact in terms of the consumer when it really belongs to the producer. This commit drops the `cook-` prefix from the FE-800 plan-emitter modules and their symbols. The orchestrator package is implicit from the directory; reframing leaves the cook command surface (`cook-cli.ts`, `runCook`, `parseCookArgs`, `CookOptions`, `brunch cook`, `.brunch/cook/plan.yaml` path) untouched. Files renamed (git mv, history preserved): - cook-plan-projection.ts → plan-projection.ts - cook-plan-llm-planning.ts → plan-llm-planning.ts - cook-plan-reconciliation.ts → plan-reconciliation.ts - cook-plan-emitter.ts → plan-emitter.ts (+ matching .test.ts files) Symbols renamed: - projectCookPlanFromSpec → projectPlanFromSpec - reconcilePlan (unchanged in name; was reconcileCookPlan) → reconcilePlan - emitCookPlanFromSnapshot → emitPlanFromSnapshot - EmitCookPlanResult → EmitPlanResult - EmitCookPlanOptions → EmitPlanOptions Memory docs (CARDS.md, PLAN.md) updated for path/symbol references and descriptive prose. Frontier id `spec-to-cook-plan` and branch name `ka/fe-800-spec-to-cook-plan` kept as stable identifiers. 285 orchestrator tests green after rename; `npm run verify` green. Amp-Thread-ID: https://ampcode.com/threads/T-019e8dea-2ba3-776f-8d0d-57c94acf5f93 Co-authored-by: Amp <amp@ampcode.com>

Slice 6: replace 'brunch plan <snapshot.json>' with 'brunch plan <specId>'. - src/server/db/completed-spec-snapshot.ts: buildCompletedSpecSnapshot(db, specId) maps accepted requirements/criteria (kind_ordinal → kindOrdinal) and active-path relationships (filtered to accepted endpoints) into the orchestrator's CompletedSpecSnapshot shape. Uses getEntitiesForSpecificationOnActivePath; orchestrator stays pure (type-only import). - src/server/plan-runner.ts: parsePlanArgs(<specId>, --out, --verbose) + runPlan({ specId, snapshot, outDir, verbose, runModel?, log? }). Header prints spec id; display rules (failure + transformation always, synthesis only with --verbose) preserved from slice 5. - src/server/cli.ts: resolves project + opens DB, calls buildCompletedSpecSnapshot then runPlan; closes DB on resolve/reject. buildCompletedSpecSnapshot is statically imported so db.ts stays bundled inline with cli.js (avoids drizzle migrations chunk-path break). - Orchestrator src/orchestrator/src/plan-cli.ts + plan-cli.test.ts deleted. - memory/PLAN.md: FE-800 status → done (all six slices). - memory/CARDS.md: retired (queue exhausted). Tests: 3 new for buildCompletedSpecSnapshot (accepted-id filter, edge filter + relation enum preservation, empty spec), 8 new for plan-runner (parsePlanArgs spec-id parsing + flags + usage errors, runPlan cycle/ synthesis/planning-failed paths). npm run verify green (1636 tests). Amp-Thread-ID: https://ampcode.com/threads/T-019e8dea-2ba3-776f-8d0d-57c94acf5f93 Co-authored-by: Amp <amp@ampcode.com>

Slice 7 (ln-review follow-up) — bundles four hardening findings on the just-landed brunch plan <specId> surface: 1. Empty-snapshot guard. CLI now rejects nonexistent specs ('specification <N> not found') and specs with no accepted requirements ('specification <N> has no accepted requirements — confirm the requirements phase before planning') before emission, instead of producing a default-epic plan that pretends to have planned a completed spec. 2. Strict arg parser. parsePlanArgs rejects unknown flags ('--bogus'), bare hyphen tokens ('-1'), --out without =, and a second positional argument. Each path throws a usage error mentioning the offending token. Existing accepted shapes unchanged. 3. Unified error boundary. The plan branch in src/server/cli.ts wraps parse + project resolve + db open + snapshot build + runPlan in one try/finally. All failure paths print the friendly 'Failed to run brunch plan: <message>' prefix and the DB is closed exactly once. 4. CLI surface oracle. Three new tests in src/server/cli.test.ts (packaged-bin suite) pin: --help mentions 'plan <specId>', 'brunch plan' with no args fails with usage error, 'brunch plan abc' fails with usage error, 'brunch plan 999' against empty .brunch/ fails with spec-not-found. 5. Lexicon normalization. Internal identifiers settle on specificationId (PlanOptions.specificationId, RunPlanArgs. specificationId, seed-helper return key). User-facing CLI token stays <specId>. Test helper variable 'project' → 'specification'. Verify green (1641 tests, 1 skipped). Live smoke: 'brunch plan 23' still emits the working plan; 'brunch plan 999' / 'brunch plan' / 'brunch plan 23 --bogus' all exit 1 with friendly messages. Amp-Thread-ID: https://ampcode.com/threads/T-019e8dea-2ba3-776f-8d0d-57c94acf5f93 Co-authored-by: Amp <amp@ampcode.com>

brunch plan <id> now writes .brunch/cook/specs/<id>/plan.yaml so multiple specs can coexist on the same project. brunch cook gains --spec=<id> and resolves plans in this precedence: 1. <dir>/plan.yaml (fixture) 2. --spec=<id> -> .brunch/cook/specs/<id>/plan.yaml (error if missing) 3. newest .brunch/cook/specs/*/plan.yaml by mtime 4. legacy .brunch/cook/plan.yaml (authored single-plan fallback) 5. otherwise error Parser rejects non-integer / non-positive --spec values. Amp-Thread-ID: https://ampcode.com/threads/T-019e8dea-2ba3-776f-8d0d-57c94acf5f93 Co-authored-by: Amp <amp@ampcode.com>

…yout Both brunch plan (writer) and brunch cook (resolver) were independently constructing .brunch/cook/specs/<id>/plan.yaml and re-implementing positive-integer validation for spec ids. Extract a small orchestrator module that owns: - specPlanPath(dir, specId) - specsRootDir(dir) - resolveLatestSpecPlanPath(dir) (newest by mtime; ignores non-int dirs) - parseSpecId(raw, flagLabel) plan-runner and cook-cli now delegate; the inline findNewestSpecPlan helper and the duplicated spec-id parser are gone. Help text under 'Plan flags' also names the spec-scoped layout. Migrated the cook-cli mtime test off execFileSync('touch', ...) to fs.utimesSync — no subprocess, no BSD/GNU date-format risk. Amp-Thread-ID: https://ampcode.com/threads/T-019e8dea-2ba3-776f-8d0d-57c94acf5f93 Co-authored-by: Amp <amp@ampcode.com>

…ition Live cook lines now read 'req-4 · users-can-drag-nodes' instead of bare 'req-4', so the operator can tell which requirement is being processed without cross-referencing plan.yaml. Display-only — slice ids stay canonical for branches, depends_on, reports, and tests. Slug rule: lowercase, cut at first clause boundary, drop stop words and sub-3-char fragments, take first 4 surviving words, cap on word boundary at 32 chars. Falls back to bare id when no usable slug emerges. Amp-Thread-ID: https://ampcode.com/threads/T-019e8dea-2ba3-776f-8d0d-57c94acf5f93 Co-authored-by: Amp <amp@ampcode.com>

…ition - slice-label.ts header example and JSDoc now match the implementation (stop words dropped anywhere, not just leading; example matches R4 smoke output). - pi-actions assess-semantic binds 'label' once like the other four actions instead of inlining the call. - plan-projection.test.ts pins the now-display-load-bearing invariant that every projected slice has a non-empty definition. Amp-Thread-ID: https://ampcode.com/threads/T-019e8dea-2ba3-776f-8d0d-57c94acf5f93 Co-authored-by: Amp <amp@ampcode.com>

kostandinang · 2026-06-04T10:38:14Z

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

This stack of pull requests is managed by Graphite. Learn more about stacking.

The first brownfield cook of spatial_graph_layout produced orphan, unwired feature code that satisfied criteria via a Ladle story, because the emitter's convention-synthesized verification.target is integration-blind. Records the integration-oracle direction (distinct from petri-simulation-oracle's net reachability) and the run-output promotion dependency as an FE-800 follow-on. Post-demo. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>

kostandinang · 2026-06-04T16:25:47Z

⚠️ Heads-up — a test is failing on this branch's tip (3d21c13c), independent of any downstack work:

src/client/__tests__/build-boundary.test.ts › "keeps rich markdown rendering lazy and preserves route-level code splitting in the production build"
→ AssertionError: expected [ …(10) ] to deeply equal [ …(10) ] (a code-splitting / chunk-set mismatch).

Confirmed it's from this branch, not downstack: it fails on fe-813 (#170) with that branch's own changes stashed, i.e. on the fe-800 tip itself. Likely introduced by a recent fe-800 commit. Since the stack inherits it, anything based on fe-800 (e.g. #170) shows red in CI until it's triaged here at the source.

cursor · 2026-06-04T16:27:12Z

PR Summary

Medium Risk
Introduces an LLM-backed planning step and changes how cook picks plan files, but deterministic reconciliation, fallbacks on LLM failure, and extensive tests limit blast radius; server→orchestrator dependency direction is intentional and one-way.

Overview
FE-800 wires the Bristol demo front half: a completed specification can become a reviewable orchestrator plan.yaml and then feed brunch cook (including --petrinaut-stream) without hand-authored plans.

New CLI and storage: brunch plan <specId> loads the project DB, builds a CompletedSpecSnapshot from accepted requirements/criteria and active-path edges (with readiness gates), runs the emitter, and writes .brunch/cook/specs/<specId>/plan.yaml. brunch cook gains --spec=<id> and expanded resolveCookMode precedence: fixture plan.yaml → explicit or newest spec plan by mtime → legacy .brunch/cook/plan.yaml. Layout and parsing live in shared spec-plan-paths.

Emitter pipeline (orchestrator): deterministic projection (requirements→slices, verifies→criterion linkage; no graph-read execution order), one-shot LLM planning (DAG, epics, non-buildable slices; injectable runModel, empty enrichment on failure), then reconciliation (cycle break, dep cleanup, unit-test targets, typed warnings). emitPlanFromSnapshot merges stages into a single EmitterWarning audit stream (transformations/failures always; synthesis only with --verbose). Orchestrator plan-cli is removed in favor of server plan-runner.

Polish and docs: cook progress logs use sliceLabel from slice definitions. memory/CARDS.md (FE-764 queue) is retired; PLAN.md, SPEC.md (A97, D160-K), and sequencing reflect FE-800 done and petri-graph-compilation reconciliation. Broad unit/integration coverage for projection, planning, reconciliation, emitter, snapshot builder, and cook resolution.

^{Reviewed by Cursor Bugbot for commit cdd876b. Bugbot is set up for automated code reviews on this repo. Configure here.}

Co-authored-by: Cursor <cursoragent@cursor.com>

cursor

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit cdd876b. Configure here.}

cursor · 2026-06-05T08:42:06Z

+    const mtimeMs = statSync(planPath).mtimeMs;
+    if (!newest || mtimeMs > newest.mtimeMs) {
+      newest = { path: planPath, mtimeMs };
+    }


Auto-pick ties break arbitrarily

Low Severity

resolveLatestSpecPlanPath only replaces the chosen plan when mtimeMs is strictly greater than the current best, so two spec plans with the same modification time leave which plan wins to readdir order. Default brunch cook resolution can pick different specs across runs on the same tree.

Additional Locations (1)

src/orchestrator/src/cook-cli.ts#L296-L297

^{Reviewed by Cursor Bugbot for commit cdd876b. Configure here.}

kostandinang and others added 10 commits June 3, 2026 15:05

kostandinang changed the title ~~FE-800: Spec to cook-plan emitter — project and plan a cook plan.yaml from a completed intent graph~~ FE-800: Spec to orchestrator plan emitter — generate plan.yaml from a completed specification Jun 4, 2026

kostandinang and others added 4 commits June 4, 2026 10:15

kostandinang mentioned this pull request Jun 4, 2026

FE-813: Cook evaluator runs the verification tests instead of an LLM verdict #170

Open

kostandinang marked this pull request as ready for review June 4, 2026 16:27

cursor Bot reviewed Jun 4, 2026

View reviewed changes

Comment thread src/server/cli.ts Outdated

kostandinang self-assigned this Jun 4, 2026

FE-800: Require confirmed criteria before planning

cdd876b

Co-authored-by: Cursor <cursoragent@cursor.com>

kostandinang requested a review from lunelson June 5, 2026 07:54

cursor Bot reviewed Jun 5, 2026

View reviewed changes

This was referenced Jun 5, 2026

FE-815: Orchestrator demo fixtures: parallel, diamond, halt #174

Draft

FE-819: Orchestrator <> Petrinaut integration improvements #176

Draft

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

FE-800: Spec to orchestrator plan emitter — generate plan.yaml from a completed specification#167

FE-800: Spec to orchestrator plan emitter — generate plan.yaml from a completed specification#167
kostandinang wants to merge 16 commits into
ka/fe-764-petri-sync-serverfrom
ka/fe-800-spec-to-cook-plan

kostandinang commented Jun 3, 2026 •

edited

Loading

Uh oh!

kostandinang commented Jun 4, 2026 •

edited

Loading

Uh oh!

kostandinang commented Jun 4, 2026

Uh oh!

cursor Bot commented Jun 4, 2026 •

edited

Loading

Uh oh!

Uh oh!

cursor Bot left a comment

Uh oh!

cursor Bot Jun 5, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

kostandinang commented Jun 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

What Changed

Scope

Verification

Uh oh!

kostandinang commented Jun 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

kostandinang commented Jun 4, 2026

Uh oh!

cursor Bot commented Jun 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

PR Summary

Uh oh!

Uh oh!

cursor Bot left a comment

Choose a reason for hiding this comment

Uh oh!

cursor Bot Jun 5, 2026

Choose a reason for hiding this comment

Auto-pick ties break arbitrarily

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

kostandinang commented Jun 3, 2026 •

edited

Loading

kostandinang commented Jun 4, 2026 •

edited

Loading

cursor Bot commented Jun 4, 2026 •

edited

Loading