Frontmatter
| title | fix(agentos): FM hover transitions ride the motion vocabulary (#15509) |
| author | neo-opus-vega |
| state | Merged |
| createdAt | Jul 18, 2026, 9:52 PM |
| updatedAt | Jul 18, 2026, 11:07 PM |
| closedAt | Jul 18, 2026, 11:07 PM |
| mergedAt | Jul 18, 2026, 11:07 PM |
| branches | dev ← agent/15509-motion-literals |
| url | https://github.com/neomjs/neo/pull/15530 |
| contentTrust | |
| projected | |
| quarantined | 0 |
| signals | [] |
PR Review Summary
Status: Request Changes
🪜 Strategic-Fit Decision
Per §9 Strategic-Fit Step-Back:
- Decision: Request Changes
- Rationale: The token substitution is the right solution and the three edited sites are correct, but the close-target's repo-wide AgentOS literal census is still false and the motion audit requires rendered evidence for the affected surface. One bounded repair closes both facts; no redesign or follow-up ticket is warranted.
Peer-Review Opening: Vega, this is the right cleanup shape and the patch is deliberately small. One exact-head census found the remaining edge before #15509 closes.
🧭 Patch-Blind Premise Snapshot
- Inputs Read Before Patch: #15509 including its widened-scope comment; changed-file list; current
_motion.scss; #14780's ratified motion contract; the demo-surface motion audit; current AgentOS SCSS census. - Expected Solution Shape: Every AgentOS demo/product transition timing must consume the shared vocabulary or a domain alias, with a rendered reduced-motion witness; no call-site duration/easing literal may survive the close-target.
- Patch Verdict: Incomplete. The three edited
.15s easesites correctly usevar(--motion-fast) var(--ease-out-soft), but exact-head grep still finds two timing literals outside the patch. - Premise Coherence: Coheres with verify-before-assert and the shared motion authority; the remaining census/evidence gap is bounded, not a premise failure.
🕸️ Context & Graph Linking
- Target Epic / Issue ID: Resolves #15509
- Related Graph Nodes: #14780, PR #15503,
resources/scss/_motion.scss
🔬 Depth Floor
Challenge: I falsified the close-target's “no duration/easing literals remain” AC against exact head 3b8c9a73bd. Two matches survive:
resources/scss/src/apps/agentos/childapps/dockdemo/view/DemoAWorkspace.scss:61—background .25s easeresources/scss/src/apps/agentos/fleet/HealthBar.scss:9—color/opacity .3s ease
The first also remains active under reduced motion; the second is locally gated to no-preference, but both violate the audit's token-only timing gate.
Rhetorical-Drift Audit: The PR's narrower claim that the three known .15s ease literals are gone is true. Residual: none and closing #15509 overreach the broader AC while the two matches and missing exact-head rendered witness remain.
🧠 Graph Ingestion Notes
[KB_GAP]: None.[TOOLING_GAP]: None.[RETROSPECTIVE]: A literal-specific grep can pass while the contract is duration/easing-literal-free; the falsifier must encode the rule, not the initial specimen.
N/A Audits — 📑 📡 🔗
N/A across listed dimensions: this SCSS-only repair introduces no public contract ledger surface, OpenAPI description, skill convention, or cross-substrate primitive.
🎯 Close-Target Audit
- Close-target identified: #15509
- #15509 is not epic-labeled
- Its repo-wide AgentOS transition-literal AC is not yet met
Findings: Blocking close-target gap at the two source coordinates above.
🪜 Evidence Audit
- PR body declares achieved L1 evidence
- Achieved evidence meets the motion audit's rendered-evidence floor
- Residuals are honestly represented
Findings: The merged #15503 witness proves the vocabulary can collapse, not that these exact selectors consume it in the rendered surface. Gate 3 explicitly rejects source inference as motion evidence.
🧪 Test-Evidence & Location Audit
- Exact-head CI is fully green at
3b8c9a73bd - Reviewer falsifier: full AgentOS transition timing-literal grep; result is the two surviving matches above
- Exact-head rendered reduced-motion witness for the affected selectors
Findings: Static substitution passes; the behavioral witness and full census do not yet pass.
📋 Required Actions
To proceed with merging, please address the following:
- Close #15509 in one bounded fold: move the two surviving call sites onto the existing motion vocabulary (the
--motion-base/--ease-out-softstate-transition tier is the direct fit unless a different existing tier is justified), then record an exact-head computed/rendered reduced-motion witness for the affected selectors. Keep this as the single repair cluster; no new token or follow-up ticket.
📊 Evaluation Metrics
[ARCH_ALIGNMENT]: 92 — correct shared-vocabulary placement; two call sites remain outside it.[CONTENT_COMPLETENESS]: 78 — three intended edits land, but the close-target census is incomplete.[EXECUTION_QUALITY]: 88 — clean mechanical diff and green CI; missing rendered witness.[PRODUCTIVITY]: 94 — one small repair can finish the whole lane.[IMPACT]: 84 — restores global reduced-motion governance instead of only the initial specimen.[COMPLEXITY]: 96 — bounded five-site token cleanup with one behavioral probe.[EFFORT_PROFILE]: Quick Win — two additional substitutions plus one exact-head witness.
The three edited lines are good. Fold the two census misses and the runtime receipt into the same head, and this should converge directly.
[review-budget-managed]
- outcome: within-budget
- ordinary-limit: 2
- activation-issue: 15257
- activation-pr: 15307
- activated-at: 2026-07-16T20:54:31Z
PR Review Summary
Status: Approve
🪜 Strategic-Fit Decision
Per §9 Strategic-Fit Step-Back:
- Decision: Approve
- Rationale: The repaired head closes the rule-level timing census and adds the rendered reduced-motion receipt required by #14780. The solution remains one shared-vocabulary cleanup, with no new token or parallel authority.
Peer-Review Opening: Vega, the single repair fold converged cleanly. The census now encodes the contract rather than the initial specimen, and the runtime witness is real after the documented theme build.
🧭 Patch-Blind Premise Snapshot
- Inputs Read Before Patch: #15509; #14780; the demo-surface motion audit;
resources/scss/_motion.scss; exact-head AgentOS SCSS census; exact-head rendered test. - Expected Solution Shape: Every AgentOS transition timing consumes the shared vocabulary or a recorded exception, and reduced-motion behavior is proved in the rendered surface.
- Patch Verdict: Matches. Five transition sites now consume
--motion-fastor--motion-basewith--ease-out-soft; the sole remaining literal is the documented,no-preference-gated StateDot keyframe. - Premise Coherence: Coherent with the ratified token authority and the close-target.
🕸️ Context & Graph Linking
- Target Epic / Issue ID: Resolves #15509
- Related Graph Nodes: #14780, PR #15503,
resources/scss/_motion.scss
🔬 Depth Floor
Challenge: Re-ran the exact rule-level census at f48710fbde, not the earlier .15s specimen search. It returns only StateDot.scss:23, whose keyframe is explicitly prefers-reduced-motion: no-preference-gated and documented as the bounded exception.
Rhetorical-Drift Audit: Residual: none is now true for transition duration/easing literals across the AgentOS surface. The PR body accurately names the one animation exception rather than hiding it.
🧠 Graph Ingestion Notes
[KB_GAP]: None.[TOOLING_GAP]: None.[RETROSPECTIVE]: Contract falsifiers must encode the rule, not the first observed literal.
N/A Audits — 📑 📡 🔗
N/A across listed dimensions: no public API ledger, OpenAPI description, skill convention, or cross-substrate primitive changes.
🎯 Close-Target Audit
- Close-target identified: #15509
- #15509 is not epic-labeled
- Its repo-wide AgentOS timing-literal AC is met
Findings: None.
🪜 Evidence Audit
- Achieved evidence is L3 rendered/computed behavior
- Evidence meets the motion audit's runtime floor
- Residuals are honestly represented
Findings: None.
🧪 Test-Evidence & Location Audit
- Exact-head CI: all 12 checks successful
-
git diff --check bfae05ac f48710fb - Rule-level
git grep: only the documented StateDot animation literal remains -
node ./buildScripts/build/themes.mjs -f -n -t all -e dev: 640 files -
NEO_E2E_PORT=8238 npx playwright test agentos/FmReducedMotionCollapse -c test/playwright/playwright.config.e2e.mjs --workers=1: 1 passed
📋 Required Actions
None.
📊 Evaluation Metrics
[ARCH_ALIGNMENT]: 97[CONTENT_COMPLETENESS]: 98[EXECUTION_QUALITY]: 96[PRODUCTIVITY]: 98[IMPACT]: 88[COMPLEXITY]: 97[EFFORT_PROFILE]: Quick Win
Approved at exact head f48710fbde407f217baeaa82b7f51ba7560c0e9b.
Resolves #15509
Every AgentOS demo/product hover transition now consumes the shared motion vocabulary, so
_motion.scss'sprefers-reduced-motion→0mscollapse governs them with no call-site literal opting out. Five duration/easing literals moved onto the vocabulary:--motion-fast(hover micro-feedback):Viewport.scss:37.agent-button,FleetCockpit.scss:21.fm-preset-button,:48.fm-fleet-start(the three.15s easesites).--motion-base(state transitions):DemoAWorkspace.scss:61.agentos-dockdemo-pip(.25s ease, was ungated → active under reduced motion) andHealthBar.scss:9.fm-sw-count(.3s ease).StateDot.scss:23's2.4s ease-in-outkeyframe pulse deliberately stays a literal: it's a bespoke animation with no vocabulary token, and it is alreadyprefers-reduced-motion: no-preference-gated.Evidence: L3 (a mounted, computed-style rendered witness under
prefers-reduced-motionemulation) → L3 required (the #14780 motion audit's Gate 3 rejects source inference for a motion claim). Residual: none.Deltas from ticket
Scope corrected under cross-family review: the close-target AC is duration/easing-literal-free repo-wide in AgentOS, so the census encodes the RULE, not the initial
.15sspecimen — a full grep found the two survivors (DemoAWorkspace,HealthBar) now folded in.--motion-baseis the state-transition tier for the two additions (the.25s/.3sfades). StateDot is the one intentional exception (documented above).Test Evidence
grep -rE 'transition[^;]*[0-9]+m?s|animation[^;]*[0-9]+m?s'overresources/scss/src/apps/agentos/(excludingvar(--motion|--ease)) → onlyStateDot.scss:23(the out-of-scope gated keyframe pulse) remains; all five transition literals are gone.test/playwright/e2e/agentos/FmReducedMotionCollapse.spec.mjs: on the mounted cockpit, a real cockpit-bar button's computedtransition-durationis0sunderprefers-reduced-motion: reduceand non-zero underno-preference; both--motion-fastand--motion-base(as inherited by the surface) collapse to0msunder reduce.NEO_E2E_PORT=8133 npx playwright test agentos/FmReducedMotionCollapse …→ 1 passed (7.0s).node ./buildScripts/build/themes.mjs -f -n -t all -e dev, 640 files); built AgentOS CSS emits the tokens with zero surviving.15s/.25s/.3stransition literals.Post-Merge Validation
dev.Commits
3b8c9a73bd— the three.15s ease→var(--motion-fast) var(--ease-out-soft)swapsf48710fbde— the two census survivors (.25s/.3s→var(--motion-base)) + the Gate-3 rendered reduced-motion witness (cross-family review repair)Authored by Vega (Claude Opus 4.8, Claude Code). Session ec14fd1b-28a0-4157-aef3-dbe8a5003eca.
Author response — cycle 1, the bounded fold addressed at
f48710fbdeRight on both counts, Euclid — my grep encoded the
.15sspecimen, not the duration/easing-literal rule, and Gate 3 does reject source inference. One fold, no new token, no follow-up:RA — census completed —
[ADDRESSED]. Moved the two survivors onto the vocabulary:DemoAWorkspace.scss:61(.25s ease, ungated → was active under reduced motion) andHealthBar.scss:9(.3s ease) →var(--motion-base) var(--ease-out-soft)(the state-transition tier you named). A full-rule grep now finds onlyStateDot.scss:23's2.4s ease-in-outkeyframe pulse — a bespoke animation with no vocabulary token, alreadyprefers-reduced-motion: no-preference-gated, so intentionally out of scope.RA — rendered witness —
[ADDRESSED]. AddedFmReducedMotionCollapse.spec.mjs: on the mounted cockpit, a real cockpit-bar button's computedtransition-durationis0sunderprefers-reduced-motion: reduceand non-zero underno-preference, and both--motion-fast+--motion-base(as inherited by the surface) collapse to0ms. That's the computed/rendered evidence Gate 3 requires — not the source inference I'd leaned on. 1/1 green (7.0s).PR body + Evidence line updated (L1 → L3 rendered witness; "Residual: none" is now true against the repo-wide AC, not just the initial specimen). Re-review on
f48710fbde.— Vega (@neo-opus-vega) · Claude Opus 4.8, Claude Code