Root causes of the remaining 5 CI PG failures (and the hidden flaky set):
1. All PG-integration packages shared ONE database under parallel package
execution - migrations/seeds/truncates raced (nondeterministic failure
sets, locally and in CI).
Fix: RequireTestDatabaseURL/SkipIfNoTestDatabaseURL now derive a
per-package database (<base>_<pkgHash>, advisory-lock-guarded
CREATE); 9 raw-env straggler tests funnelled through the helpers;
db-package tests call it unqualified (no import cycle).
2. Approval-chain fixtures were not production-shaped: freezeReviewSnapshot
only honors head_snapshot_id, fake content hashes never froze a
snapshot, so ApproverApprove rejected with 'approved requires frozen
review_snapshot_id'. Errors were swallowed (_, _ =) hiding this.
Fix (ownership-respecting, arch-enforced):
- docrevision.SeedContentSnapshotForTest seeds the ledger tables it owns
- docgov GovernanceEngine.FreezeHeadSnapshotForTest freezes head/review
pointers on document_records (docgov-owned) and clears engine cache
- all four integration fixtures use the pair; swallowed errors checked
3. Fixed-name fixtures collide once a package shares one DB (projects
uq_projects_name_global): uniquified in projectinstance + server tests.
4. CAS/ctxmgr tests never bootstrapped migrations (relied on other packages
having migrated the shared DB): explicit RunMigrationsOnceForTest.
Verified: full parallel short suite = 75 packages, 0 failures against an
isolated throwaway PG. CI skip list removed - short tests fully required.
Runtime verification found the deterministic generator merges the prior
served narrative, so a replaced (semantic) narrative would linger after
rollback. Store the displaced deterministic narrative in the shadow report
(persisted via shadow_json) and merge that instead when the prior summary
source is semantic_replace.
Unseal SEMANTIC_SUMMARY_REPLACE under explicit user authorization with a
defense-in-depth session allowlist (SEMANTIC_SUMMARY_REPLACE_SESSIONS):
swap happens only for allowlisted sessions after result digests and served
watermark validation, bounded narrative size; facts/counts stay deterministic
and the daily digest stays advisory. Master flag alone replaces nothing.
User authorized lifting CURRENT FAV whole-branch merge protection on
2026-09-23. Keep FAV_DEGRADED_READ/FAV_PUBLISHER_ENABLED default false.
Drop duplicate goose 00034 (local_dev already has 00036). Preserve
agentx workspace admin routes and pin-export wiring.
Close T20-B2/B6/B8/B9/B10 low-risk leaks: map-backed cancel reclaim, idempotency lock TTL in cleanupLoop, sync.Once information_schema probe, InvalidateLayer after chat memory upsert, and Subscribe before SSE headers.
arch-check --mode block failed on frozen dev (23 NEW import findings):
modules materialindex/mineru/workflowscope and edges shipped via their
own accepted tasks were never registered in module-boundaries.yaml.
Register the intentional allowlist edges (precedent 97b868a9); no new
exceptions are added.
7 PG tests fell back to a hardcoded dev DB (127.0.0.1:5433) when
TEST_DATABASE_URL/CONTEXT_DB_DSN were unset, contradicting the isolated
test-DB contract (BUG-08) and polluting dev on a bare 'go test -short'.
Change the fallback to skip when no isolated DB env is provided.
Co-authored-by: Cursor <cursoragent@cursor.com>
A new recheck mark is posted into the actor's existing project session, and a downstream-impact event does the same. Asking about what a change affects still resolves to impact analysis.
Co-authored-by: Cursor <cursoragent@cursor.com>
Absorb stalled capability fixes and avoid goose out-of-order on schema 57 DBs by shipping chat_summaries as 00058.
Co-authored-by: Cursor <cursoragent@cursor.com>
Deny foreign user_id upsert without manage rights (403); unique project
names in ctxmgr PG fixture for global name index; add test-env data repair SQL.
Co-authored-by: Cursor <cursoragent@cursor.com>
Shadow evaluates a no-tool summarizer on caller-supplied messages; served
narrative stays deterministic and REPLACE remains unwired.
Co-authored-by: Cursor <cursoragent@cursor.com>
Extend unapplied migration to mark 6001–6010 sessions demo_retired
and filter them (and retired projects) out of ListChatSessions.
Co-authored-by: Cursor <cursoragent@cursor.com>
Keep ctxmgr as table owner and filter the authorized visible set in-process so window, stats, restore, digest, and prompt assembly share one Principal-scoped collection without a package move.
Co-authored-by: Cursor <cursoragent@cursor.com>
CAS the unique terminal status with the assistant turn, frozen manifest, envelope and FAV staging, and join ctxmgr writes through the same UoW instead of a process mutex.
Co-authored-by: Cursor <cursoragent@cursor.com>
Approve/commit now write FAV staging in the domain transaction, recover-apply materializes scratch collections, and unpublished or GET paths cannot mutate the registry.
Co-authored-by: Cursor <cursoragent@cursor.com>
Approve and index only ledger blobs; task success now completes the chat turn and recomputes export columns.
Co-authored-by: Cursor <cursoragent@cursor.com>
Persist workflow_instance_id, stage_instance_id, scope_kind on chat
messages and task runs; resolve scope from live instance for chat and
document tasks so refresh can apply consistent stage authorization.
Co-authored-by: Cursor <cursoragent@cursor.com>
Tag messages with stage_key metadata, block sends for unauthorized roles, and filter history/search/case results by allowed stages.
Co-authored-by: Cursor <cursoragent@cursor.com>
Extend taskrun executors to return full TaskRunResult, finalize assistant
placeholder content with artifact metadata, and map quality_failed terminal
state instead of metadata-only updates.
Co-authored-by: Cursor <cursoragent@cursor.com>
agent-bot
weichen