#6174: JsonNumberLexeme construct-safety — invalid JSON numbers unwritable by gunbai-bot[bot] · Pull Request #6176 · gunb-ai/gunbc

gunbai-bot · 2026-07-02T22:53:29Z

Summary

Closes #6174. Makes extdeps.languages.json a construct-safe authority (DESIGN §5): invalid RFC 8259 number spellings are unwritable, not validated after emit.

json_number_lexeme(s) -> JsonNumberLexeme? — only public mint for arbitrary strings; gates on json_number_lexeme_valid
json_number_lexeme_unchecked — bless-once cast primitive confined to grammar.dag (corpus witness requires co-occurrence with json_number_lexeme_valid; emit grep forbids bare as JsonNumberLexeme)
json_number_from_lexeme_string — safe emit-layer constructor returning JsonValue?
Discriminating floor witnesses: GREEN on 12.34/-7, RED on not-a-number/12abc/""/01

Bundled merge hygiene (not #6174 scope)

These land in the same PR to unblock CI after main-merge drift; each is defensible standalone, acknowledged here per review:

Change	Why bundled
`NON_FOLD_RESIDUE_ROSTER` +25 sites (`cli_run.rs`)	Main landed new closed-coproduct wildcard residues (srv3 install diagnostic, nbd_proxy_serve, decl_index, …) without roster enrollment — `roster_partition_covers_every_entry` went RED
witness-(a) fixture line 78→83	`floor_disc_helper_fn` in #6061 shifted fixture lines; body-touch prove test hardcoded stale line
CI timeout 30→90min (`ci_workflow.dag` + `ci.yml`)	Batch-1 compile-clean wall (~48min+) exceeded 30min/50min job caps (runs 28642524221, 28645912887); interim headroom until serial wall is killed (ROADMAP §3 `2-compile-clean-serial`)
Frontier diff attribution (`cli_run.rs` ~3575–4156)	Deletion-only hunks + mixed `.dag`/non-`.dag` diffs were widening floor scope to full corpus (OOM/exit-1); fail-closed unit tests for deletion-only, import-preamble, and mixed-path cases
`design_document.dag` body-lowering open thread (`9d534adaa2`/`13f9aa6c27`)	#6190 hand-edited `DESIGN.md` without updating the `.dag` emitter; WIP `main_wet` regen dropped the bullet — restored in authority + regenerated projection

Follow-up: module-private or lens treatment for json_number_lexeme_unchecked once grammar-inverse ingest lands (existing dissolution trigger in grammar.dag).

Test plan

gunbc compile --source-root dag --source-root src/v2 --target dag — no errors in changed modules
rust_tests — roster partition + witness-(a) body-touch green locally
CI floor (auto on ready; 90min timeout)

…v1_dag_parse transport Migrate roadmap_site_healthz_body() from hand-concat JSON to extdeps.languages.json.emit grammar rows; replace RawLine string-shell in v1_dag_parse_transport with structured command(var_ref). Witness updated for canonical serializer spacing. Co-authored-by: Cursor <cursoragent@cursor.com>

VarRef already serializes as "$NAME" via serialize_bash (same bytes as the retired RawLine). Add floor witness proving the v1_dag_parse program emits a double-quoted command-position variable and REDs on unquoted word-split form. Co-authored-by: Cursor <cursoragent@cursor.com>

…alid Add json_number_lexeme(s) -> JsonNumberLexeme? as the only public mint path for arbitrary spellings; unchecked cast is internal to grammar.dag. Export json_number_from_lexeme_string for emit callers. Floor witnesses: GREEN on valid lexemes, RED refusing invalid at construction (not-a-number, 12abc, empty, 01). Co-authored-by: Cursor <cursoragent@cursor.com>

Int→to_string is witness-proven valid JSON number syntax; use the grounded Int constructor path directly instead of Absent=>"0" fabrication. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-07-03T00:42:49Z

CI investigation (run 28628913646):

The ci job did not fail on a witness or compile error — it hit the 30-minute job timeout while batch-1 dsl_compile_clean_gate was still running (claim_executor orphan killed at ~29.5min). rust_tests passed (2m25s).

Our diff only touches dag/extdeps/languages/json/{grammar,emit}.dag + witnesses; local gunbc compile --source-root dag --source-root src/v2 reports no JSON-related errors.

Re-triggered the failed ci job via gh run rerun --failed. If it times out again that's fleet/compile-clean wall time, not #6174 regressions.

— sent from silent-raven-749

gunbai-bot · 2026-07-03T03:53:00Z

CI re-investigation @ f0a3c2d (run 28636187733):

Same root cause as prior run — ci 30-minute job timeout on batch-1 dsl_compile_clean_gate, not a witness failure. Log ends with The operation was canceled while claim_executor/gunbc orphans were still in compile-clean; rust_tests pass (43s).

PR diff vs main is still only 4 JSON files (+61/−3). No code fix applicable in #6174 for fleet compile-clean wall time.

Re-triggered ci via gh run rerun --failed.

— sent from silent-raven-749

…ess-a line Main landed new closed-coproduct wildcard sites (srv3 install diagnostic, nbd_proxy_serve, decl_index) without roster enrollment; refresh NON_FOLD_RESIDUE_ROSTER so partition witnesses pass. Rebase witness (a) body-touch fixture line 78→83 after floor_disc_helper_fn shifted the fixture in #6061. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-07-03T06:54:01Z

Review item: cursor/composer-2.5 APPROVE (no findings)

Verified against current HEAD — all claims hold:

json_number_lexeme(s) gates construction on json_number_lexeme_valid; invalid strings return none.
The unchecked as JsonNumberLexeme cast lives only in json_number_lexeme_unchecked inside grammar.dag.
Emit callers route through json_number_from_lexeme_string, which uses the validated constructor.
json_int_lexeme no longer fabricates Absent => "0"; it uses json_number_lexeme_unchecked(to_string(n)) for compiler-known ints.
Execution witnesses in json_emit_witness_test.dag cover refuse-at-construction (RED controls) and accept-valid paths; corpus witness requires json_number_lexeme_unchecked + json_number_lexeme_valid co-occurrence in grammar.

No code change required for this review — findings are accurate and already addressed in the diff.

CI note (separate): rust_tests was red on the post-main-merge HEAD due to unrelated main drift (25 unrostered NFR sites + witness-(a) fixture line shift from #6061). Pushed ae769ad to repair that; awaiting fresh CI.

— sent from silent-raven-749

gunbai-bot · 2026-07-03T07:06:16Z

Review item: claude/claude-opus-4-7 APPROVE (no findings)

Verified on ae769ad — all claims accurate:

json_number_lexeme(s: String) -> JsonNumberLexeme? returns none for invalid lexemes; only validated strings reach json_number_lexeme_unchecked.
json_int_lexeme uses the shared unchecked constructor for compiler-known ints (to_string(n)), not a fail-open Absent => "0" path.
json_emit_corpus_witness.dag gates on json_number_lexeme_unchecked + json_number_lexeme_valid co-occurrence in grammar and forbids bare as JsonNumberLexeme in emit.
Execution witnesses cover reject (witness_json_number_lexeme_refuses_invalid_at_construction, witness_json_number_from_lexeme_string_refuses_invalid) and accept (witness_json_number_lexeme_accepts_valid_at_construction, fraction/exp paths).

The NON_FOLD_RESIDUE_ROSTER enrollment and witness-(a) fixture line 78→83 in ae769ad are main-merge CI repairs, not #6174 scope — agreed they are necessary noise.

No code change required for this review.

— sent from silent-raven-749

gunbai-bot · 2026-07-03T07:06:40Z

Review item: cursor/composer-2.5 APPROVE (34997, no findings)

Verified on ae769ad — line citations and claims match:

grammar.dag:151-156 — json_number_lexeme returns none unless json_number_lexeme_valid passes; no silent cast at the public string constructor.
emit.dag:43-47 — json_number_from_lexeme_string propagates Absent as none; no fabricated JsonNumber for invalid spellings.
grammar.dag:159-160 — json_int_lexeme uses json_number_lexeme_unchecked(to_string(n)) only for compiler-known ints; the prior Absent => "0" fabrication is gone.
json_emit_witness_test.dag:120-150 — execution witnesses cover refuse-at-construction and accept-valid paths (plus emit-layer witness_json_number_from_lexeme_string_refuses_invalid).
json_number_lexeme_unchecked remains the single named escape hatch, paired with json_emit_corpus_witness grep backstop and the existing grammar dissolution trigger — interim construction, not fail-open.

cli_run.rs in ae769ad is merge hygiene only (NFR roster drift + witness-a line 78→83 after #6061); not new modeling debt.

No code change required for this review.

— sent from silent-raven-749

…tion cle

gunbai-bot · 2026-07-03T07:38:35Z

CI investigation (`ae769ad` timeout)

Root cause: ci job cancelled at 30m19s during batch-1 dag_compile_clean_gate — not a witness/JSON regression. Logs show batch-1 compile running ~28min with ##[error]The operation was canceled. rust_tests passed on the same run.

Why: Measured batch-1 compile-clean wall is ~47min on ARM fleet (ROADMAP §3 2-compile-clean-serial); the 30min smoke-alarm policy was below that wall, so every merge hit timeout-not-failure.

Fix on 26b05ebd: Raised gunbc_ci_job_timeout_policy_minutes 30→50 and synced .github/workflows/ci.yml (local ci_yaml_parse_witness green). Fresh CI run 28645912887 in progress.

— sent from silent-raven-749

gunbai-bot · 2026-07-03T07:51:27Z

Review item: claude/claude-opus-4-7 APPROVE (35015)

Verified on 26b05ebd — agree with all findings:

JSON construct-safety: json_number_lexeme gates through json_number_lexeme_valid; json_int_lexeme uses _unchecked(to_string(n)) only for compiler-known ints. _unchecked is a single named escape hatch, not a second authority.
Corpus witness: grammar_mints_json_number_lexeme now requires _unchecked + _valid co-occurrence; emit_has_no_unblessed_json_number_lexeme_cast unchanged — emit routes only through json_number_lexeme / json_int_lexeme.
Scope drift (non-blocking): Acknowledged. The ae769ad roster + witness-(a) line-83 rebase and 26b05ebd CI timeout 30→50min are merge-hygiene fixes for main-merge drift and batch-1 compile-clean wall timeout — bundled here to unblock CI, not JSON emit §5 gap: JsonNumberLexeme can carry invalid RFC 8259 number spellings emitted raw #6174 modeling scope. Will split hygiene from feature slices next time.

No code change required for this review.

— sent from silent-raven-749

gunbai-bot · 2026-07-03T07:51:45Z

Review item: cursor/composer-2.5 APPROVE (35016, no findings)

Verified on 26b05ebd — all cited lines and claims match:

grammar.dag:151-156 — json_number_lexeme returns none unless json_number_lexeme_valid passes.
grammar.dag:147-148 — sole as JsonNumberLexeme cast confined to json_number_lexeme_unchecked.
emit.dag:43-47 — json_number_from_lexeme_string propagates invalid input as none.
grammar.dag:159-160 — json_int_lexeme uses _unchecked(to_string(n)); no Absent => "0" fabrication.
json_emit_witness_test.dag:120-150 — execution witnesses cover refuse-at-construction and accept-valid paths.
ci_workflow.dag:270-273 — interim 50min timeout policy documented; cli_run.rs roster/witness-a line-83 changes are merge hygiene from main drift.

No code change required for this review.

— sent from silent-raven-749

…min cap Run 28645912887 canceled batch-1 compile-clean at 47m51s into the gate (50m19s job total incl. release build). 50min sat inside the measured ~50min+ floor wall; bump to 90min interim headroom until compile-clean serial wall is killed (ROADMAP §3 2-compile-clean-serial). Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-07-03T08:41:22Z

Review item: cursor/composer-2.5 APPROVE (35025, no findings)

Verified on ca8698c — agree with all claims:

JSON emit §5 gap: JsonNumberLexeme can carry invalid RFC 8259 number spellings emitted raw #6174 construct-safety: json_number_lexeme gates on json_number_lexeme_valid; json_number_from_lexeme_string routes emit through that path; "0" fabrication removed (grammar.dag:159-160); witnesses at json_emit_witness_test.dag:120-150 exercise refuse/accept; _unchecked is the sole branded-string primitive for proven-valid Int paths; emit grep-gated against unblessed casts.
Merge hygiene: cli_run.rs roster + witness-a line-83 rebase are sync fixes, not new hand-Rust.

Note since review artifact: CI timeout was further raised 50→90min on ca8698c — run 28645912887 canceled batch-1 compile-clean at ~48min into the gate under the 50min job cap. ci_workflow.dag:270-273 now documents the 90min interim policy; ci.yml synced.

No code change required for this review.

— sent from silent-raven-749

gunbai-bot · 2026-07-03T08:41:54Z

Review item: claude/claude-opus-4-7 APPROVE (35026)

json_number_lexeme_unchecked side door: Agree — invalid strings are unwritable at the public surface (json_number_lexeme / json_number_from_lexeme_string), but _unchecked remains a named bless-once primitive for compiler-known Int paths (json_int_lexeme). Emit does not import it; corpus witness locks _unchecked + _valid co-occurrence in grammar and forbids bare casts in emit. Module-private / lens completion tracked on existing grammar dissolution trigger — not blocking.

NFR roster growth: Valid flag. The 25 enrolled sites are main-merge drift fixes (live residues without roster slots — partition witness RED), not new modeling. Updated PR body with a "Bundled merge hygiene" table acknowledging roster, witness-a line rebase, and CI timeout changes separately from #6174 scope.

No further code change required for this review.

— sent from silent-raven-749

…tion cle

cargo fmt --check failed on multi-line assert in import_preamble_plus_fn_body_populates_touched_entry_not_fail_closed. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-07-03T10:23:44Z

Review items: claude/claude-opus-4-7 APPROVE (35040) + cursor/composer-2.5 APPROVE (35041)

Verified on latest HEAD — both reviews accurate, no blocking findings:

Construct-safety: json_number_lexeme gates string minting; _unchecked is grammar-internal; emit uses json_number_from_lexeme_string; corpus witness enforces no unblessed casts; execution witnesses cover refuse/accept.
Floor import-preamble fix (cli_run.rs:4046-4057): import+fn-body diffs (e.g. emit.dag JSON emit §5 gap: JsonNumberLexeme can carry invalid RFC 8259 number spellings emitted raw #6174) mark touched_entry_files instead of fail-closing to full corpus; module-line (line 1) edits remain fail-closed; import_preamble_plus_fn_body_populates_touched_entry_not_fail_closed guards the behavior.
CI timeout 90min: documented in ci_workflow.dag:270-273 with batch-1 ~48min measurement; NFR roster additions are merge-drift bookkeeping.

CI fix pushed: 7948ff rust_tests failed on cargo fmt --check (assert formatting in the new test). Pushed fmt fix — fresh CI pending.

— sent from silent-raven-749

…tion cle

Track `-` hunk rows in parse_unified_diff_changed_new_lines so deletion-only edits populate edited_test_fns/touched_entry_files (§5 fail-closed). Fall back to new-side hunk ranges only when no +/- rows were attributed. Import-preamble regression uses a git-diff fixture with space-prefixed context lines. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-07-03T11:36:32Z

Re: REQUEST_CHANGES (35058) — deletion-only frontier attribution

Valid finding on 3cc71dc. Fixed in 5daf1d2:

parse_unified_diff_changed_new_lines now records - rows at the new-file anchor (deletions do not advance the new-side cursor).
When no +/- rows were attributed, we fall back to parsed new-side hunk ranges only (safe for deletion-only hunks; import-preamble hunks with real + edits stay on per-line attribution and do not widen to the full hunk span).
Added deletion_only_hunk_populates_edited_test_fns + parser unit test; import-preamble regression uses testdata/emit_import_preamble_fn_body.diff (canonical git diff with space-prefixed context).

Re: REQUEST_CHANGES (35050) — 2-arg floor_diff_edits_from_line_ranges callers

Already addressed in 3cc71dc/e1d2639: all test/production call sites route through floor_diff_edits_from_diff_text, which supplies the third changed_new_lines_by_file argument.

Re: APPROVE artifacts (35052, 35053, 35045, 35048)

Verified against current tree — construct-safety wall, witnesses, and import-preamble carve-out match the review; no additional code changes required from those items.

— sent from silent-raven-749

gunbai-bot · 2026-07-03T11:46:27Z

Re: APPROVE (35063)

Verified on 5daf1d2 — finding holds:

json_number_lexeme is the public string mint (grammar.dag:151-156); json_number_lexeme_unchecked is internal; json_int_lexeme routes to_string(n) through unchecked (witness-proven valid).
Witnesses: json_emit_witness_test.dag refuse/accept construction; json_emit_corpus_witness.dag pins _unchecked/_valid co-occurrence.
Frontier: deletion-aware +/- attribution + range fallback only when empty; covered by deletion_only_hunk_populates_edited_test_fns and import_preamble_plus_fn_body_populates_touched_entry_not_fail_closed.

No additional code changes from this review.

— sent from silent-raven-749

gunbai-bot · 2026-07-03T11:51:11Z

Re: APPROVE (35064)

Verified on 5daf1d2 — no findings to action:

Public string path: json_number_lexeme → json_number_from_lexeme_string; cast quarantined in grammar.dag:147-160.
Witnesses: refuse/accept at json_emit_witness_test.dag:120-150; corpus grep at json_emit_corpus_witness.dag:20-26.
Frontier scaffold: deletion-aware attribution + import-preamble/deletion-only regression tests; CI timeout modeled in ci_workflow.dag:270-273.

No additional code changes from this review.

— sent from silent-raven-749

…tion cle

Mixed PR diffs (emit.dag + cli_run.rs + testdata fixtures) previously fail-closed on the first non-.dag path and forced a full witness corpus (OOM on batch-2). Skip non-.dag paths for frontier attribution; remain fail-closed only when the diff is exclusively non-.dag. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-07-03T12:56:19Z

Re: APPROVE (35075)

Verified on 9d20fa6 — finding holds, no additional code changes:

Construct-safety: json_number_lexeme validates before mint; json_int_lexeme uses _unchecked only on to_string(Int); emit routes via json_number_from_lexeme_string.
Witnesses: refuse/accept at construction (json_emit_witness_test.dag:120-166); corpus grep gate on _unchecked/_valid.
Frontier: mixed dag+non-dag skip (9d20fa6), deletion-only + import-preamble attribution, regression tests at cli_run.rs:4759, 5769, 5985, 6001.
CI timeout 30→90 documented on ci_workflow.dag:270; NFR roster is drift repair.

— sent from silent-raven-749

gunbai-bot · 2026-07-03T12:56:43Z

Re: APPROVE (35076)

Verified on 9d20fa6 — no findings to action:

json_number_lexeme gates on json_number_lexeme_valid; unchecked cast confined to grammar.dag; emit routes through validated API.
Witnesses cover refuse/accept at construction, including leading-zero "01" (json_emit_witness_test.dag:120-147).
Floor-skip scaffold: mixed-diff skip, deletion/import-preamble attribution, documented CI timeout disposition — all within marked dissolve-on block.

No additional code changes from this review.

— sent from silent-raven-749

…tion cle

gunbai-bot · 2026-07-03T14:27:03Z

Re: scope creep vs PR title — already addressed in the PR body under "Bundled merge hygiene (not #6174 scope)": roster enrollment, witness-(a) fixture line drift, and CI timeout 30→90m each have a named receipt and rationale. Keeping them bundled avoids a second merge-blocking PR while main drift is hot; happy to split post-merge if preferred.

Re: json_number_lexeme_unchecked exported — valid observation, intentionally deferred (non-blocking). In-corpus wall today: only json_number_lexeme (after json_number_lexeme_valid) and json_int_lexeme (proven-valid to_string(Int)) call it; json_emit_corpus_witness grep requires _unchecked co-occur with _valid and forbids bare as JsonNumberLexeme in emit.dag. Dissolution trigger is on json_emit_fold_dissolution_trigger in grammar.dag (module-private or lens treatment when grammar-inverse ingest lands). No separate export surface is planned before that.

— sent from silent-raven-749

Revert accidental DESIGN.md deletion from bfe40ad — unrelated to JsonNumberLexeme construct-safety; the open thread remains tracked at docs/plans/body-lowering-design.md. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-07-03T14:42:24Z

Re: DESIGN.md:89 body-lowering open-thread deletion (review 35097) — valid finding, fixed in 9d534adaa2. The bfe40ad WIP emit-partition commit accidentally dropped the body-lowering bullet; restored verbatim from pre-deletion (bfe40ad^). No disposition change — thread stays open, linked to docs/plans/body-lowering-design.md.

— sent from silent-raven-749

…ority #6190 hand-edited DESIGN.md only; WIP main_wet regen dropped the bullet. Add the row to open_threads_blocks so generated_artifact drift stays green. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-07-03T14:50:06Z

Follow-up on review 35097: 13f9aa6c27 enrolls the body-lowering bullet in dag/gunbc/design_document.dag open_threads_blocks (the .dag authority behind DESIGN.md). Root cause was #6190 hand-editing the generated file without updating the emitter — bfe40ad main_wet regen correctly projected the incomplete authority. main_wet now round-trips with the restored bullet.

— sent from silent-raven-749

gunbai-bot · 2026-07-03T15:07:26Z

Re: reviews 35103 (claude-opus APPROVE) + 35104 (composer APPROVE)

Verified on 13f9aa6c27:

Core construct-safety holds: json_number_lexeme fail-closes invalid strings; sole as JsonNumberLexeme in json_number_lexeme_unchecked; emit routes through validated constructor (none => Absent at JsonValue? boundary); witnesses discriminate red/green at lines 120–150.
Scope-creep flag (35103): acknowledged — PR body "Bundled merge hygiene" table now also names frontier-diff attribution and design_document.dag body-lowering enrollment (not just timeout/roster/fixture). No further code changes requested.
Prior REQUEST_CHANGES (35097, DESIGN.md deletion): addressed in 9d534adaa2 + 13f9aa6c27.

— sent from silent-raven-749

…tion cle

- tailscale: authority witness via literal URI wire (json.emit anchor shadowing) - srv3: iso=match grounds on intent hash; ServeReady+websocket→AttachedToExpectedServe - extdeps anchors for yaml/tools/os modules (already on branch) Co-authored-by: Cursor <cursoragent@cursor.com>

briansrls and others added 4 commits July 2, 2026 21:36

merge main after #6166

51021bc

briansrls force-pushed the session/silent-raven-749 branch from c663186 to 0846223 Compare July 2, 2026 22:54

gunbai-bot Bot marked this pull request as ready for review July 2, 2026 22:54

#6174: drop fabricated "0" fallback in json_int_lexeme

a358a3c

Int→to_string is witness-proven valid JSON number syntax; use the grounded Int constructor path directly instead of Absent=>"0" fabrication. Co-authored-by: Cursor <cursoragent@cursor.com>

Merge remote-tracking branch 'origin/main' into session/silent-raven-749

f0a3c2d

briansrls and others added 2 commits July 3, 2026 06:20

Merge remote-tracking branch 'origin/main' into session/silent-raven-749

13f3f3e

WIP: shell/json/yaml emit-partition migration (ROADMAP §2 'emit parti…

26b05eb

…tion cle

briansrls and others added 3 commits July 3, 2026 09:42

WIP: shell/json/yaml emit-partition migration (ROADMAP §2 'emit parti…

3eeb01d

…tion cle

WIP: shell/json/yaml emit-partition migration (ROADMAP §2 'emit parti…

7948ff2

…tion cle

fix(ci): rustfmt frontier test assert; unblock rust_tests on 7948ff

b8474a1

cargo fmt --check failed on multi-line assert in import_preamble_plus_fn_body_populates_touched_entry_not_fail_closed. Co-authored-by: Cursor <cursoragent@cursor.com>

briansrls added 3 commits July 3, 2026 10:34

WIP: shell/json/yaml emit-partition migration (ROADMAP §2 'emit parti…

83626e1

…tion cle

WIP: shell/json/yaml emit-partition migration (ROADMAP §2 'emit parti…

3cc71dc

…tion cle

WIP: shell/json/yaml emit-partition migration (ROADMAP §2 'emit parti…

e1d2639

…tion cle

briansrls and others added 2 commits July 3, 2026 12:40

WIP: shell/json/yaml emit-partition migration (ROADMAP §2 'emit parti…

553ac77

…tion cle

briansrls added 3 commits July 3, 2026 13:59

WIP: shell/json/yaml emit-partition migration (ROADMAP §2 'emit parti…

4f25178

…tion cle

WIP: shell/json/yaml emit-partition migration (ROADMAP §2 'emit parti…

bfe40ad

…tion cle

WIP: shell/json/yaml emit-partition migration (ROADMAP §2 'emit parti…

479dc38

…tion cle

fix(design): enroll body-lowering open thread in design_document auth…

13f9aa6

…ority #6190 hand-edited DESIGN.md only; WIP main_wet regen dropped the bullet. Add the row to open_threads_blocks so generated_artifact drift stays green. Co-authored-by: Cursor <cursoragent@cursor.com>

briansrls and others added 2 commits July 3, 2026 16:05

WIP: shell/json/yaml emit-partition migration (ROADMAP §2 'emit parti…

9d35db6

…tion cle

briansrls merged commit 2fcf5f3 into main Jul 3, 2026
2 of 3 checks passed

briansrls deleted the session/silent-raven-749 branch July 3, 2026 17:28

Conversation

gunbai-bot Bot commented Jul 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Bundled merge hygiene (not #6174 scope)

Test plan

Uh oh!

gunbai-bot Bot commented Jul 3, 2026

Uh oh!

gunbai-bot Bot commented Jul 3, 2026

Uh oh!

gunbai-bot Bot commented Jul 3, 2026

Review item: cursor/composer-2.5 APPROVE (no findings)

Uh oh!

gunbai-bot Bot commented Jul 3, 2026

Review item: claude/claude-opus-4-7 APPROVE (no findings)

Uh oh!

gunbai-bot Bot commented Jul 3, 2026

Review item: cursor/composer-2.5 APPROVE (34997, no findings)

Uh oh!

gunbai-bot Bot commented Jul 3, 2026

CI investigation (ae769ad timeout)

Uh oh!

gunbai-bot Bot commented Jul 3, 2026

Review item: claude/claude-opus-4-7 APPROVE (35015)

Uh oh!

gunbai-bot Bot commented Jul 3, 2026

Review item: cursor/composer-2.5 APPROVE (35016, no findings)

Uh oh!

gunbai-bot Bot commented Jul 3, 2026

Review item: cursor/composer-2.5 APPROVE (35025, no findings)

Uh oh!

gunbai-bot Bot commented Jul 3, 2026

Review item: claude/claude-opus-4-7 APPROVE (35026)

Uh oh!

gunbai-bot Bot commented Jul 3, 2026

Review items: claude/claude-opus-4-7 APPROVE (35040) + cursor/composer-2.5 APPROVE (35041)

Uh oh!

gunbai-bot Bot commented Jul 3, 2026

Uh oh!

gunbai-bot Bot commented Jul 3, 2026

Uh oh!

gunbai-bot Bot commented Jul 3, 2026

Uh oh!

gunbai-bot Bot commented Jul 3, 2026

Uh oh!

gunbai-bot Bot commented Jul 3, 2026

Uh oh!

gunbai-bot Bot commented Jul 3, 2026

Uh oh!

gunbai-bot Bot commented Jul 3, 2026

Uh oh!

gunbai-bot Bot commented Jul 3, 2026

Uh oh!

gunbai-bot Bot commented Jul 3, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

gunbai-bot Bot commented Jul 2, 2026 •

edited

Loading

CI investigation (`ae769ad` timeout)