Batch fix for review-visual-flow's two P0 + five P1 findings across the
Chat / Rubrics / Growth surfaces. Everything routes through tokens — no
new hex family, no gradient hero, no violet competing with --accent.
P0-1 blue collision — plan-signal chip shared the exact --accent hue
with --turn-action-edge, so a .sig-plan chip inside an action turn fused
visually with the turn's left rail. Introduces --signal-plan-fg / -bg /
-border (+ restart variants) on a teal axis (light #0f766e, dark
#5eead4), rewires .turn-signal-chip.sig-plan, .sig-plan-restart, their
dark override, and the .trace-timeline-signal-badge / .trace-graph-
signal-ring SVG variants to the tokens. Now the four signal families
(error red / loop red / redundant amber / plan teal) all sit off the
blue axis so the turn-edge blue keeps its exclusive whole-turn-action
semantic.
P0-2 rubric-hint-card hero — was a full-width purple gradient banner
(12px radius, 20px icon, primary CTA) violating spec §2 'L1 material
styled as L0 hero'. Repainted as a 28px compact row: single hairline on
--surface, 4px radius, downgraded icon to '·', inlined title+subtitle,
CTA to ghost small. Reads as a Fields-tree-grade nudge, not page chrome.
P1 batch (5 CSS token conversions):
1. .rubric-tile-stats-rate/-dot .pass/.fail: raw #79d17b/#f96e6e →
--status-ok-fg / --status-err-fg.
2. .rubrics-empty-illust svg [fill]: raw #a89a83 → --icon-mute.
3. .growth-filter-chip.active: raw --accent-* → --nav-item-active-*.
4. .kb-tabs .active hue: raw --accent-fg → --text.
5. .rubric-grid-name size: 15px → 14px (three-size rule).
7 renderer test file (visual-consistency-polish.test.js) locks all the
above via style.css AST assertions. 3 shots + 3 before shots under
docs/qa-visual-fix/ document the delta on chat / rubrics / growth.
10 files, +400/-49. Visual-only, no behavioral changes.
The cancel-tool-calls scenario cancels a live bash call, which relies on
POSIX detached-process-group termination; bash has no Windows process-tree
kill yet (deferred with the Bash execution domain), so the hung call times
the scenario out on the native Windows snapshot lane.
Add a posixOnly scenario declaration that skips the run test on win32
while the fixture guards keep covering committed files on every platform,
and mark cancel-tool-calls with it.
The two transcript tests that assert footer token counters inherited
process.cwd() as the session cwd. In a checkout deep enough that the
footer label exceeds the 88-column fake terminal, the counters never
render and the assertions fail. Pin those tests to a short fixed cwd;
cwd rendering keeps its dedicated variants test.
Declare the tool-result pruning plugin in the examples workspace so the repl Cordis configuration resolves through plain Node and the Loader metadata gate.
Express the validated single-node surface rewrite without a non-null assertion or an unreachable defensive branch, preserving both the runtime contract and per-file 100% coverage.
Retarget the pruning feature onto the merged compaction foundation.
Adapt content-only tool-result rewrite validation to the session-owned surface manager, migrate the demo wiring to repl-agent, and refresh generated type, catalog, and website contracts while preserving pressure and overflow pruning behavior.
The childToolOmissions comparison arms ran only through the examples
snapshot suites, which the unit coverage gate does not count. Add an
authored child-omission scenario to the synthetic replay suite: one
scripted child omits the declared tool (header pin minus the omission,
prompt pin skipped), one keeps the full set (pin and prompt compared
verbatim), restoring 100% branch coverage on suite.ts.
An omitted maxDepth meant unbounded recursion, and the shipped examples
shipped that default. maxDepth now defaults to 1; a numeric cap requires
the provider's depthLimit capability (the mount fails loud and points to
the explicit 'provider-managed' opt-out for out-of-process providers),
and a child AT the cap loses the delegating tool from its own schema via
the child toolFilter — prompt-face hiding on top of the execution-face
depth check. Examples pin maxDepth explicitly. The ACP snapshot harness
gains Scenario.childToolOmissions so a child session may legitimately
omit declared delegation tools from its pinned header and prompt;
affected subagent/workflow goldens are re-recorded.
A subagent child's recursion depth lived only in runtime AgentOptions,
so a persisted child came back from resume counted as top-level and
maxDepth stopped binding after every restart. Add
SessionHeader.delegationDepth, round-trip it through the JSONL and
SQLite backends (SQLite schema v5), restore it on agent-loop resume,
and write it when the in-process backends create a child. The seam now
owns the shared depth vocabulary (delegationDepthOf): the persisted
header is authoritative and monotone — runtime options may deepen it
but never lower it.
Adopt initiator-derived private loop and session plumbing while preserving post-step recovery and balanced synthetic tool-call results after cancellation.
Keep the architecture map within its existing budget while documenting ambient Agent ownership and explicit cross-boundary identities.
Combine agent initiator scoping and exact Cordis JSDoc inspection with post-step compaction and bounded request recovery.
Regenerate the Cordis and website API catalogs, and classify the recovery and compaction types required by the current catalog link-coverage gate.
ZiyaZhang
Yichen Jiang
Tianyi Cui
imccyu