Why: task-sync merged authoritative main (card-only change); the evidence
gate flagged prior rows as predating non-card changes, so the combo make
test was rerun at the synced head and re-recorded.
Evidence: evidence row on this card (SELF, head ddc8963).
Impact: task card only.
Refs: rollout I4
Why: include the current integration baseline before verification.
Evidence: fetched origin/main=682a4e8263fe007cb102fc9deaf71bda514ee387.
Impact: integrate upstream changes without rewriting task history.
Refs: task-sync
Why: combo make test against frontend pin 8b59998 (Frontend #104 merged)
is green; evidence row tool-written before opening the Dev PR.
Evidence: evidence row on this card (SELF).
Impact: task card only.
Refs: rollout I4
Why: the canonical e2e.sh run (frontend c9ba344 + duty-based seed) passed
13/13 plus plain-base 2/2 — the I4 acceptance scenarios are green and the
evidence row is tool-written.
Evidence: evidence row on this card (SELF, sha256 in record).
Impact: task card only.
Refs: rollout I4
Why: vite dev server (4174) login returns 401 with credentials that succeed
via direct curl to the same serve - a separate proxy-forwarding anomaly in
the dev-server path, recorded so the next session uses the canonical
e2e.sh-managed preview instead.
Evidence: p10 probe (401 via dev proxy) vs direct curl 200 on the same
serve/data, same credentials.
Impact: card bookkeeping.
Refs: rollout I4
Why: the instrumented run (render-log build) confirms the DOM-vs-state
contradiction and adds the instance-remount hypothesis with the missing
done-log evidence.
Evidence: probe logs show RENDER/EFFECT/FETCH-resolved complete but
setProjects-done absent, select=0 at t+3s and t+8s.
Impact: card bookkeeping.
Refs: rollout I4
Why: the cockpit hang is now isolated to the React commit layer - the
component's fetch resolves (rows=2, cancelled=false, setProjects called)
but the DOM keeps rendering the loading state. This contradiction needs
interactive React DevTools work; documented with the reproduction path.
Evidence: trace.zip (run 20261016-160406 and the 013103 preview rerun);
console capture shows resolved rows=2 while the DOM snapshot shows 加载中.
Impact: card bookkeeping; no code change this commit.
Refs: rollout I4
Why: multiple 加载中 hangs traced to preview restarts without
PLAYWRIGHT_API_TARGET (proxy landing on the shared dev RDM at 8787);
recorded so the next session verifies the preview environment first.
Evidence: manual probes with/without the env var (login 401 vs entry
visible) on the same seed data.
Impact: card bookkeeping only.
Refs: rollout I4
Why: the seven-scenario suite is merged but scenario 1 hangs at 加载中 in the
preview environment; trace evidence shows the component fetch resolving
(rows=2, cancelled=false) while the DOM stays on the loading state - an
unresolved state-vs-DOM contradiction needing interactive debugging in a
stable environment. Recorded per honest-reporting rules instead of merging
a red suite.
Evidence: trace.zip of run 20261016-160406 (console: effect start, resolved
rows=2; DOM: 加载中); existing suites 8/8 green in the same run.
Impact: I4 browser acceptance remains open; no product regression - the
same run's audit suites pass.
Refs: rollout I4
Why: card must record the gitlink follow-up to frontend main head ffda996
(sidebar wiring, payload unwrap, login settle) via a fresh successful
transport after a failed-receipt reuse mistake.
Evidence: run success transport 29c76d31..., apply a555bc6c (SELF).
Impact: card bookkeeping only.
Why: card must carry the full worker run history: r1 decode miss, r2 bare
JSON success, r3/r4 block placement fix with transient TLS between, r5
doc.write grant after the adopt danger-gate discovery (resolved via RDM
#426 structure-content, gate kept intact), and the pin transport with one
correctly-blocked contradictory packet.
Evidence: receipts in /tmp; applies 0a1e777/b1ba22d2/372dca1/b013be1e/
27ead8f7.
Impact: card bookkeeping only.
Why: card must record the block-placement fix (live run proved the CLI
structure write must sit in the stop_serve window) and the transient TLS
failure between successful transports.
Evidence: r3 receipt (97-byte TLS log, no work started); r4 receipt
(success, transport de88901c..., applied as b1ba22d2); placement checks.
Impact: card bookkeeping only.
Why: card must record r1 failed-unknown (decode miss, site preserved) and
r2 success transport per agent-exec governance.
Evidence: receipts /tmp/dev-e2e-i4-receipt.json and
/tmp/dev-e2e-i4-r2-receipt.json; apply 0a1e777.
Impact: card bookkeeping only.
Why: establish the authorized task workspace and its initial baseline.
Evidence: preserved precheck receipt in .task/feature-document-visualization-cockpit-i4.md; R0=aa3c7eae40264ee47d5505a162074cfd77b1a099.
Impact: task metadata only; the feature workspace starts at aa3c7eae40.
Refs: .task/feature-document-visualization-cockpit-i4.md
weichen
agent-exec-transport