You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
vitals-review: run log for skilld-dev/skilld.dev #679
2026-10-31T06:21:00.000Z — vitals-review on skilld-dev/skilld.dev | 0 found | 0 new | 0 already known | 0 outside allowed scope | 0 over 5 files | 0 issues requested
Site: s_08aae654, https://skilld.dev. Matched exactly against tracked nuxt.config.ts → site.url.
Field coverage: Cloudflare returned one Poor finding. CrUX summary and trend are unavailable, so origin p75 direction is unknown.
Lab coverage: 16 pages, 32 strategy groups, 100 retained Scans. Judged 99 completed Scans; excluded one failed Scan. Read FCP details for the latest two completed Scans of each eligible group, plus reference details. History is bounded by the retained limit.
No new fix ready to file. Poor measurements remain.
No persistent Lab drop was established. No Suspect or MissingMetric verdict arose in the judged suffixes. Older FCP onset remains undated where only the latest Poor pair was needed.
Excluded failed Scan: 075a50cd-0c8c-4650-9065-bf51a87a114a, basecamp mobile, September 25. Its null metrics were not treated as zeros.
Poor lab dispositions
Homepage LCP: latest mobile values are 5101 and 5851 ms. Both name a demo image. Observed resource load delays are approximately 3598 and 4967 ms. These observed phases do not partition modeled LCP.
app/pages/_HomeDemoCarousel.vue shuffles the current demo during onMounted. PR #576 explicitly requests a random starting order per visit. Changing that behavior requires Harlan’s decision about preserving variety without replacing the initial image after hydration. No proposal overrides that decision.
Shared CSS and FCP: action 01M2J4SG3QK0095X6TVB0BP6T9 names blocking entry and component stylesheets. PR #500 already restored inline component styles and recorded remaining homepage and Skill LCP cost. Current configuration retains that fix. Live homepage HTML still links entry, DemoWriting, and WhyVisual CSS.
No distinct new CSS fix was established. Harlan owns remaining-cost triage. The other Poor pages share this evidence, so they receive no separate proposals. Several retained Skill and org Scans also measure missing-page content after redirects; their measurements do not establish a new route repair.
Fix follow-up and coverage limits
No closed routine:vitals-review issues were returned. No routine-owned fix qualified for a post-deploy Scan.
Public repository visibility was verified. Open issues, open PR bodies, matching closed issues, and relevant merged PRs were read. No matching open performance repair was found.
Site status is degraded because health.crawlBlockerCount is unavailable. Its Next Action concerns indexing, outside this routine. Lighthouse versions vary across retained history, limiting causal comparisons.
Repository and GitHub stayed read only. No NuxtSEO write was made.
Confidence: 88/100. Untested: local causal replay and field recovery after today’s deploy.
2026-11-04T01:00:00.000Z — vitals-review on skilld-dev/skilld.dev | 0 found | 0 new | 0 already known | 0 outside allowed scope | 0 over 5 files | 0 issues requested
Site: s_08aae654, https://skilld.dev. Exactly matched sites list against tracked nuxt.config.ts → site.url.
Field coverage: unavailable. CrUX summary and trend report insufficient traffic. Origin p75 direction is unknown. Web Analytics findings report source: null, no field element data, and total: 0. This does not establish good performance.
Lab coverage: 20 pages, 40 separate page/strategy series. Read 100 retained Scans, including 99 completed and one failed. Read 76 successful Scan details for FCP and cause evidence.
No proposals returned. Poor lab measurements exist, but current cause confirmation and duplicate checks remain incomplete.
Lab verdicts
Each cell lists LCP / FCP / TBT / CLS. Short means ShortSeries. Counts cover the contiguous latest Lighthouse release only.
No Lab drop was established. The mobile TBT Suspect and desktop TBT Unstable series file nothing.
Version boundaries excluded older measurements for homepage mobile; both nuxt-seo strategies; mobile torch_npu, code-review and szeyu; desktop theory-of-constraints, omarchy, vercel-ai-sdk and tristanmanchester; and mobile better-chatbot-patterns. Single Scans after version changes prove neither persistence nor recovery.
FCP details establish the latest-pair verdicts above. Earlier FCP details were intentionally bounded for Poor series. They do not establish a complete FCP history or an earlier onset.
Skipped failed Scan: 075a50cd-0c8c-4650-9065-bf51a87a114a, mobile omarchy, 2026-09-25. Its null metrics were excluded.
Poor measurements and disposition
Pairs below run from older to newer. Values are milliseconds.
Withheld: current cause and ownership checks incomplete
/skills/api/git/openclaw-multi-search-engine
4972 → 5791
4620 → 4533
Withheld: current cause and ownership checks incomplete
/orgs/tristanmanchester
Not persistently Poor
3696 → 3241
Withheld: current cause and ownership checks incomplete
The four Skill pages map through legacy redirects to /gh routes and layers/registry/app/components/SkillDetail.vue. Retired /orgs routes map through layers/registry/server/utils/orgs-route-policy.ts to author profiles.
Performance action 01M2J4SG3QK0095X6TVB0BP6T9 identifies shared render-blocking CSS. Its producer evidence is aged, dated 2026-10-14. The detail response exposes 50 of 51 evidence items, with a continuation cursor.
Scan details identify blocking entry CSS, SkillDetail CSS and DemoWriting CSS on the affected Skill pages. SkillDetail.vue statically imports _SkillDemo.vue, which imports _DemoWriting.vue. This provides a possible shared cause, but does not prove the change needed to remove the measured delay.
Live read-only checks followed the api/git and omarchy redirects. Both responses still link three stylesheets. Their entry asset is now entry.UM25vn7w.css, while the latest Poor Scans cite entry.CvwryD2V.css. Current configuration also contains the component-style build change from commit 80d9b79c, PR #500. These observations do not establish current lab persistence or confirm a fix outcome.
No page was selected as the highest-traffic representative of this shared cause. No separate proposals were created for its affected pages or metrics.
Ownership and fix follow-up
The controller supplied complete open issue and pull request lists. Most ownership bodies were complete. Issue #655 and PR #680 bodies were truncated.
For each of the five Poor page series above, the missing duplicate check is the complete body comparison against #655 and #680. Their visible dependency scope does not establish ownership of the CSS defect. It also does not justify claiming complete duplicate checks. No existing owner was positively matched.
Closed issue history was absent. Matching prior fixes, closing pull requests and their deployments could not be checked. Fix follow-up is incomplete. No fix was declared confirmed, failing or recurrent.
Next actor: controller, supply complete ownership bodies and closed routine history. A later review also needs current cause evidence before proposing a shared CSS fix.
All required Skills were present. Repository and GitHub access remained read only. No NuxtSEO mutation ran. Evidence is retained under /home/agent/scratch/vitals-review/20261104T113600Z/.
Confidence: 85/100. Current performance impact, complete duplicate checks and prior-fix deployment outcomes remain unverified.
No Lab drop was established. ShortSeries can also mean insufficient reference Scans, despite a measured increase.
Only the latest contiguous Lighthouse release was compared. Version 13.6.0 leaves one Scan for each homepage strategy. It also leaves single-Scan windows for several routes. Older 13.5.0 measurements cannot establish persistence across that boundary.
FCP details cover the latest two completed Scans where persistence can be tested. Older details were read where comparisons required them. Single-Scan windows remain short; their missing FCP details establish no recovery.
Performance action 01M2J4SG3QK0095X6TVB0BP6T9 attributes blocking time to older entry and component CSS. Its source clock is 2026-10-14. Scan details also show redirects and, on several routes, missing-Skill screens.
Repository mapping reaches nuxt.config.ts, SkillDetail.vue, and the legacy redirect middleware. Live reads confirm the four Poor LCP URLs redirect to /gh/.... Omarchy's current HTML still links entry, SkillDetail, and _DemoWriting stylesheets, with changed hashes. Changed hashes prove neither reduced blocking cost nor recovery.
The current configuration already contains style-handling work from commit 80d9b79c, PR #500. Another generic inline-styles proposal would lack a verified distinct cause. No specific fix within five files was established.
Ownership and fix follow-up
Controller open issue and PR lists are complete. Their supplied bodies were checked for matching work.
2026-11-18T01:00:00.000Z — vitals-review on skilld-dev/skilld.dev | 1 found | 1 new | 0 already known | 0 outside allowed scope | 0 over 5 files | 1 issues requested
Site: s_08aae654, https://skilld.dev. Exactly one Site matches tracked nuxt.config.ts → site.url.
Field coverage: no element findings. source: null. CrUX summary and trend both return available: false. Origin p75 direction is unknown.
Lab coverage: 100 retained completed Scans, 23 pages, 46 separate strategy series. No failed Scans skipped. Read FCP details for all 83 Scans within the latest contiguous Lighthouse version windows.
One repository proposal: preserve the server-rendered homepage hero during hydration.
app/pages/_HomeDemoCarousel.vue shuffles all demos in onMounted, including the selected opening image. This replaces the image the browser already discovered. The latest mobile Scan identifies a carousel screenshot as LCP. Keeping the opening demo stable removes that second image selection and request during startup. The source is unchanged between this checkout and controller deployment commit 48986bf8cd16108ba51c897aa0c13c38ebbd5416.
The latest two homepage mobile LCP measurements are Poor: 5101 and 4351 ms. The preceding same-version measurement is 3901 ms. This establishes a Poor series, rather than a Drop with sufficient reference history.
Performance action 01M2J4SG3QK0095X6TVB0BP6T9 also names blocking stylesheets. Latest Scan details still list component CSS. The earlier inline-style repair exists in commit 80d9b79c (#500). Its closed issue and deployment evidence are unavailable, so this run makes no fix-outcome claim. The carousel proposal addresses a separately verified cause.
Lab verdicts follow. Values in each strategy cell are ordered LCP / FCP / TBT / CLS. Short means ShortSeries. Short-ref means ShortSeries because fewer than two reference measurements exist. Neither files work.
Version boundaries: Lighthouse 13.5.0 → 13.6.0 excludes 17 older Scans across eight groups. These are both homepage strategies, both /gh/harlan-zw/nuxt-seo strategies, mobile torch_npu, desktop theory-of-constraints, mobile code-review, and mobile /orgs/szeyu. Only one compatible Scan remains in five of those groups. Older matching measurements were never joined across a version boundary.
The four older Skill pages with persistent Poor LCP/FCP now return 301 redirects. Checked api/git, basecamp, fal3, and jackspace. Their original template measurements do not establish a current repository repair. /orgs/tristanmanchester also returns 301 to /@tristanmanchester; its Poor FCP series does not justify changing the retired template. No proposals for those stale route measurements.
Field dispositions: no rows returned. Missing field coverage does not establish good user performance.
Ownership coverage: controller open issue and PR lists are complete. Complete bodies show no matching carousel repair. PR #695 has a truncated dependency-update body, so its omitted content remains an ownership coverage gap for the homepage proposal. No complete duplicate-check claim is made for that omitted content.
Closed routine issues were not supplied. Fix follow-up is incomplete, including any matching closed carousel work. No prior fix outcome is declared. Controller deployment history covers recent November 18 runs only. It cannot identify deploys between the November 4 reference Scan and November 11 first Poor Scan. The commit range remains unavailable. The controller must supply that historical evidence; no credentials were requested.
No post-deploy Scan was started: no eligible merged-fix follow-up was established, and the viewer credential lacks page:write.
Confidence: 85/100. Carousel replacement is verified in source; its LCP benefit still needs a controlled local before/after measurement.
Controller decisions
Lab mobile LCP on /: filed as vitals-lab#/#mobile#lcp, Poor for the last 2 measured Scans since 2026-11-11T03:39:36.000Z. Latest 4351 ms. Last better Scan: 2026-11-04T03:15:25.000Z.
The carousel shuffles its entire demo order in onMounted while picked remains zero. This can replace the server-rendered opening image during hydration, forcing another image request before the final hero can paint. Latest mobile Scan details identify a carousel screenshot as LCP. Preserve the server-selected opening demo and defer any reordering until navigation. This removes the startup image replacement. Performance action 01M2J4SG3QK0095X6TVB0BP6T9 describes additional blocking CSS, which this focused repair does not claim to remove. Source matches controller deployment commit 48986bf. Complete open ownership bodies contain no matching repair; PR #695's omitted body and closed-fix history remain coverage gaps.
Lab evidence: mobile Poor for the last 2 measured Scans since 2026-11-11T03:39:36.000Z. Latest 4351 ms. Last better Scan: 2026-11-04T03:15:25.000Z. Suspect commits: unavailable: controller deployment history does not cover 2026-11-04T03:15:25.000Z..2026-11-11T03:39:36.000Z.
Lab verification: run Unlighthouse on / with --mobile before and after the fix.
Activity
harlan-github-agent commentedon Oct 31, 2026
2026-10-28T01:00:00.000Z — Skipped. This run was due more than 6 hours ago, so it was skipped.
harlan-github-agent commentedon Oct 31, 2026
2026-10-31T06:21:00.000Z — vitals-review on skilld-dev/skilld.dev | 0 found | 0 new | 0 already known | 0 outside allowed scope | 0 over 5 files | 0 issues requested
Site:
s_08aae654,https://skilld.dev. Matched exactly against trackednuxt.config.ts→site.url.Field coverage: Cloudflare returned one Poor finding. CrUX summary and trend are unavailable, so origin p75 direction is unknown.
Lab coverage: 16 pages, 32 strategy groups, 100 retained Scans. Judged 99 completed Scans; excluded one failed Scan. Read FCP details for the latest two completed Scans of each eligible group, plus reference details. History is bounded by the retained limit.
No new fix ready to file. Poor measurements remain.
Full report and diagram
Field disposition
/,#hero-heading: LCP p75 4532 ms, Poor, 210 estimated views, Cloudflare window October 4–31. Latest mobile lab CLS: 0.048.The finding qualifies by severity and traffic. A distinct new fix was not established:
private, no-storedoes not establish an edge cache failure.786a28a1, finishing at 05:30:41 UTC.The field heading and lab demo image are different LCP elements. Their causes cannot be substituted.
Every lab series
Each cell lists LCP / FCP / TBT / CLS.
P = Poor series; S = Steady; U = Unstable; Short = ShortSeries for all four metrics.
//gh/harlan-zw/nuxt-seo/skills/api/git/openclaw-multi-search-engine/skills/basecamp/omarchy/omarchy/skills/fal3/claude-skills-collection/swift-modern-architecture-skill/skills/jackspace/claudeskillz/better-chatbot-patterns/skills/existential-birds/beagle/vercel-ai-sdk/skills/finos/morphir-dotnet/elm-to-fsharp-guru/skills/ascend-ai-coding/awesome-ascend-skills/torch_npu/skills/andurilcode/skills/theory-of-constraints/skills/amorriscode/agent-grimoire/code-review/orgs/tristanmanchester/orgs/harlan-zw/orgs/jaredstewart/orgs/szeyu/orgs/g1cloudNo persistent Lab drop was established. No Suspect or MissingMetric verdict arose in the judged suffixes. Older FCP onset remains undated where only the latest Poor pair was needed.
Excluded failed Scan:
075a50cd-0c8c-4650-9065-bf51a87a114a, basecamp mobile, September 25. Its null metrics were not treated as zeros.Poor lab dispositions
Homepage LCP: latest mobile values are 5101 and 5851 ms. Both name a demo image. Observed resource load delays are approximately 3598 and 4967 ms. These observed phases do not partition modeled LCP.
app/pages/_HomeDemoCarousel.vueshuffles the current demo duringonMounted. PR #576 explicitly requests a random starting order per visit. Changing that behavior requires Harlan’s decision about preserving variety without replacing the initial image after hydration. No proposal overrides that decision.Shared CSS and FCP: action
01M2J4SG3QK0095X6TVB0BP6T9names blocking entry and component stylesheets. PR #500 already restored inline component styles and recorded remaining homepage and Skill LCP cost. Current configuration retains that fix. Live homepage HTML still links entry, DemoWriting, and WhyVisual CSS.No distinct new CSS fix was established. Harlan owns remaining-cost triage. The other Poor pages share this evidence, so they receive no separate proposals. Several retained Skill and org Scans also measure missing-page content after redirects; their measurements do not establish a new route repair.
Fix follow-up and coverage limits
routine:vitals-reviewissues were returned. No routine-owned fix qualified for a post-deploy Scan.Public repository visibility was verified. Open issues, open PR bodies, matching closed issues, and relevant merged PRs were read. No matching open performance repair was found.
Site status is degraded because
health.crawlBlockerCountis unavailable. Its Next Action concerns indexing, outside this routine. Lighthouse versions vary across retained history, limiting causal comparisons.Repository and GitHub stayed read only. No NuxtSEO write was made.
Confidence: 88/100. Untested: local causal replay and field recovery after today’s deploy.
harlan-github-agent commentedon Nov 4, 2026
2026-11-04T01:00:00.000Z — vitals-review on skilld-dev/skilld.dev | 0 found | 0 new | 0 already known | 0 outside allowed scope | 0 over 5 files | 0 issues requested
s_08aae654,https://skilld.dev. Exactly matchedsites listagainst trackednuxt.config.ts→site.url.source: null, no field element data, andtotal: 0. This does not establish good performance.No proposals returned. Poor lab measurements exist, but current cause confirmation and duplicate checks remain incomplete.
Lab verdicts
Each cell lists LCP / FCP / TBT / CLS.
Shortmeans ShortSeries. Counts cover the contiguous latest Lighthouse release only.//gh/harlan-zw/nuxt-seo/orgs/tristanmanchester/orgs/g1cloud/orgs/harlan-zw/orgs/jaredstewart/orgs/szeyu/skills/jackspace/claudeskillz/better-chatbot-patterns/skills/existential-birds/beagle/vercel-ai-sdk/skills/basecamp/omarchy/omarchy/skills/fal3/claude-skills-collection/swift-modern-architecture-skill/skills/finos/morphir-dotnet/elm-to-fsharp-guru/skills/amorriscode/agent-grimoire/code-review/skills/andurilcode/skills/theory-of-constraints/skills/api/git/openclaw-multi-search-engine/skills/ascend-ai-coding/awesome-ascend-skills/torch_npu/skills/atemndobs/nebula-rfp/convex-patterns/skills/camera-2018/astrbot-plugin-dev-skill/astrbot-plugin-dev/skills/data-fair/lib/data-fair-ws/skills/dawiddutoit/custom-claude/test-property-basedNo Lab drop was established. The mobile TBT Suspect and desktop TBT Unstable series file nothing.
Version boundaries excluded older measurements for homepage mobile; both nuxt-seo strategies; mobile torch_npu, code-review and szeyu; desktop theory-of-constraints, omarchy, vercel-ai-sdk and tristanmanchester; and mobile better-chatbot-patterns. Single Scans after version changes prove neither persistence nor recovery.
FCP details establish the latest-pair verdicts above. Earlier FCP details were intentionally bounded for Poor series. They do not establish a complete FCP history or an earlier onset.
Skipped failed Scan:
075a50cd-0c8c-4650-9065-bf51a87a114a, mobile omarchy, 2026-09-25. Its null metrics were excluded.Poor measurements and disposition
Pairs below run from older to newer. Values are milliseconds.
/skills/jackspace/claudeskillz/better-chatbot-patterns/skills/basecamp/omarchy/omarchy/skills/fal3/claude-skills-collection/swift-modern-architecture-skill/skills/api/git/openclaw-multi-search-engine/orgs/tristanmanchesterThe four Skill pages map through legacy redirects to
/ghroutes andlayers/registry/app/components/SkillDetail.vue. Retired/orgsroutes map throughlayers/registry/server/utils/orgs-route-policy.tsto author profiles.Performance action
01M2J4SG3QK0095X6TVB0BP6T9identifies shared render-blocking CSS. Its producer evidence is aged, dated 2026-10-14. The detail response exposes 50 of 51 evidence items, with a continuation cursor.Scan details identify blocking entry CSS, SkillDetail CSS and DemoWriting CSS on the affected Skill pages.
SkillDetail.vuestatically imports_SkillDemo.vue, which imports_DemoWriting.vue. This provides a possible shared cause, but does not prove the change needed to remove the measured delay.Live read-only checks followed the api/git and omarchy redirects. Both responses still link three stylesheets. Their entry asset is now
entry.UM25vn7w.css, while the latest Poor Scans citeentry.CvwryD2V.css. Current configuration also contains the component-style build change from commit80d9b79c, PR #500. These observations do not establish current lab persistence or confirm a fix outcome.No page was selected as the highest-traffic representative of this shared cause. No separate proposals were created for its affected pages or metrics.
Ownership and fix follow-up
The controller supplied complete open issue and pull request lists. Most ownership bodies were complete. Issue #655 and PR #680 bodies were truncated.
For each of the five Poor page series above, the missing duplicate check is the complete body comparison against #655 and #680. Their visible dependency scope does not establish ownership of the CSS defect. It also does not justify claiming complete duplicate checks. No existing owner was positively matched.
Closed issue history was absent. Matching prior fixes, closing pull requests and their deployments could not be checked. Fix follow-up is incomplete. No fix was declared confirmed, failing or recurrent.
Next actor: controller, supply complete ownership bodies and closed routine history. A later review also needs current cause evidence before proposing a shared CSS fix.
All required Skills were present. Repository and GitHub access remained read only. No NuxtSEO mutation ran. Evidence is retained under
/home/agent/scratch/vitals-review/20261104T113600Z/.Confidence: 85/100. Current performance impact, complete duplicate checks and prior-fix deployment outcomes remain unverified.
harlan-github-agent commentedon Nov 11, 2026
2026-11-11T01:00:00.000Z — vitals-review on skilld-dev/skilld.dev | 0 found | 0 new | 0 already known | 0 outside allowed scope | 0 over 5 files | 0 issues requested
s_08aae654,https://skilld.dev. Matched exactly through trackednuxt.config.ts→site.url.null. CrUX summary and trend are unavailable. Origin p75 direction is unknown.No repair proposal is ready. Poor measurements persist, but a specific current repository fix remains unverified.
Full review note
Field disposition
vitals findingsreports: “No field element data.” No field rows qualify for return. This is missing coverage, not evidence of good performance.Lab verdicts
Each cell lists LCP / FCP / TBT / CLS.
S: SteadyP: Poor seriesShort: ShortSeriesU: UnstableSus: SuspectPaths below retain their measured
/skillsand/orgsprefixes. Strategies were judged separately.//gh/harlan-zw/nuxt-seo/orgs/g1cloud/orgs/harlan-zw/orgs/jaredstewart/orgs/szeyu/orgs/tristanmanchester/skills/amorriscode/agent-grimoire/code-review/skills/andurilcode/skills/theory-of-constraints/skills/api/git/openclaw-multi-search-engine/skills/ascend-ai-coding/awesome-ascend-skills/torch_npu/skills/atemndobs/nebula-rfp/convex-patterns/skills/basecamp/omarchy/omarchy/skills/camera-2018/astrbot-plugin-dev-skill/astrbot-plugin-dev/skills/data-fair/lib/data-fair-ws/skills/dawiddutoit/custom-claude/test-property-based/skills/existential-birds/beagle/vercel-ai-sdk/skills/fal3/claude-skills-collection/swift-modern-architecture-skill/skills/finos/morphir-dotnet/elm-to-fsharp-guru/skills/frankxai/arcanea/reasoningbank%20intelligence/skills/fredm00n/framerlabs/framer-plugins/skills/jackspace/claudeskillz/better-chatbot-patterns/skills/simonlee2/claude-plugins/seedance-video-genNo Lab drop was established. ShortSeries can also mean insufficient reference Scans, despite a measured increase.
Only the latest contiguous Lighthouse release was compared. Version
13.6.0leaves one Scan for each homepage strategy. It also leaves single-Scan windows for several routes. Older13.5.0measurements cannot establish persistence across that boundary.FCP details cover the latest two completed Scans where persistence can be tested. Older details were read where comparisons required them. Single-Scan windows remain short; their missing FCP details establish no recovery.
Poor measurements and disposition
api/git/openclaw-multi-search-enginebasecamp/omarchy/omarchyfal3/claude-skills-collection/swift-modern-architecture-skilljackspace/claudeskillz/better-chatbot-patterns/orgs/tristanmanchesterPerformance action
01M2J4SG3QK0095X6TVB0BP6T9attributes blocking time to older entry and component CSS. Its source clock is 2026-10-14. Scan details also show redirects and, on several routes, missing-Skill screens.Repository mapping reaches
nuxt.config.ts,SkillDetail.vue, and the legacy redirect middleware. Live reads confirm the four Poor LCP URLs redirect to/gh/.... Omarchy's current HTML still links entry, SkillDetail, and _DemoWriting stylesheets, with changed hashes. Changed hashes prove neither reduced blocking cost nor recovery.The current configuration already contains style-handling work from commit
80d9b79c, PR #500. Another generic inline-styles proposal would lack a verified distinct cause. No specific fix within five files was established.Ownership and fix follow-up
All required CLI reads succeeded. Final git status was clean. Repository and GitHub access stayed read only.
Confidence: 84/100. Current lab cost, field experience, and closed-fix outcomes remain unverified.
harlan-github-agent commentedon Nov 18, 2026
2026-11-18T01:00:00.000Z — vitals-review on skilld-dev/skilld.dev | 1 found | 1 new | 0 already known | 0 outside allowed scope | 0 over 5 files | 1 issues requested
s_08aae654,https://skilld.dev. Exactly one Site matches trackednuxt.config.ts→site.url.source: null. CrUX summary and trend both returnavailable: false. Origin p75 direction is unknown.One repository proposal: preserve the server-rendered homepage hero during hydration.
app/pages/_HomeDemoCarousel.vueshuffles all demos inonMounted, including the selected opening image. This replaces the image the browser already discovered. The latest mobile Scan identifies a carousel screenshot as LCP. Keeping the opening demo stable removes that second image selection and request during startup. The source is unchanged between this checkout and controller deployment commit48986bf8cd16108ba51c897aa0c13c38ebbd5416.The latest two homepage mobile LCP measurements are Poor: 5101 and 4351 ms. The preceding same-version measurement is 3901 ms. This establishes a Poor series, rather than a Drop with sufficient reference history.
Performance action
01M2J4SG3QK0095X6TVB0BP6T9also names blocking stylesheets. Latest Scan details still list component CSS. The earlier inline-style repair exists in commit80d9b79c(#500). Its closed issue and deployment evidence are unavailable, so this run makes no fix-outcome claim. The carousel proposal addresses a separately verified cause.Lab verdicts follow. Values in each strategy cell are ordered LCP / FCP / TBT / CLS.
Shortmeans ShortSeries.Short-refmeans ShortSeries because fewer than two reference measurements exist. Neither files work.//gh/harlan-zw/nuxt-seo/skills/simonlee2/claude-plugins/seedance-video-gen/skills/fredm00n/framerlabs/framer-plugins/skills/frankxai/arcanea/reasoningbank%20intelligence/skills/dawiddutoit/custom-claude/test-property-based/skills/atemndobs/nebula-rfp/convex-patterns/skills/data-fair/lib/data-fair-ws/skills/ascend-ai-coding/awesome-ascend-skills/torch_npu/skills/camera-2018/astrbot-plugin-dev-skill/astrbot-plugin-dev/skills/andurilcode/skills/theory-of-constraints/skills/amorriscode/agent-grimoire/code-review/orgs/szeyu/skills/api/git/openclaw-multi-search-engine/skills/basecamp/omarchy/omarchy/orgs/harlan-zw/orgs/jaredstewart/skills/fal3/claude-skills-collection/swift-modern-architecture-skill/skills/finos/morphir-dotnet/elm-to-fsharp-guru/skills/existential-birds/beagle/vercel-ai-sdk/orgs/g1cloud/skills/jackspace/claudeskillz/better-chatbot-patterns/orgs/tristanmanchesterVersion boundaries: Lighthouse 13.5.0 → 13.6.0 excludes 17 older Scans across eight groups. These are both homepage strategies, both
/gh/harlan-zw/nuxt-seostrategies, mobiletorch_npu, desktoptheory-of-constraints, mobilecode-review, and mobile/orgs/szeyu. Only one compatible Scan remains in five of those groups. Older matching measurements were never joined across a version boundary.The four older Skill pages with persistent Poor LCP/FCP now return 301 redirects. Checked
api/git,basecamp,fal3, andjackspace. Their original template measurements do not establish a current repository repair./orgs/tristanmanchesteralso returns 301 to/@tristanmanchester; its Poor FCP series does not justify changing the retired template. No proposals for those stale route measurements.Field dispositions: no rows returned. Missing field coverage does not establish good user performance.
Ownership coverage: controller open issue and PR lists are complete. Complete bodies show no matching carousel repair. PR #695 has a truncated dependency-update body, so its omitted content remains an ownership coverage gap for the homepage proposal. No complete duplicate-check claim is made for that omitted content.
Closed routine issues were not supplied. Fix follow-up is incomplete, including any matching closed carousel work. No prior fix outcome is declared. Controller deployment history covers recent November 18 runs only. It cannot identify deploys between the November 4 reference Scan and November 11 first Poor Scan. The commit range remains unavailable. The controller must supply that historical evidence; no credentials were requested.
No post-deploy Scan was started: no eligible merged-fix follow-up was established, and the viewer credential lacks
page:write.Confidence: 85/100. Carousel replacement is verified in source; its LCP benefit still needs a controlled local before/after measurement.
Controller decisions
The carousel shuffles its entire demo order in onMounted while picked remains zero. This can replace the server-rendered opening image during hydration, forcing another image request before the final hero can paint. Latest mobile Scan details identify a carousel screenshot as LCP. Preserve the server-selected opening demo and defer any reordering until navigation. This removes the startup image replacement. Performance action 01M2J4SG3QK0095X6TVB0BP6T9 describes additional blocking CSS, which this focused repair does not claim to remove. Source matches controller deployment commit 48986bf. Complete open ownership bodies contain no matching repair; PR #695's omitted body and closed-fix history remain coverage gaps.
Lab evidence: mobile Poor for the last 2 measured Scans since 2026-11-11T03:39:36.000Z. Latest 4351 ms. Last better Scan: 2026-11-04T03:15:25.000Z. Suspect commits: unavailable: controller deployment history does not cover 2026-11-04T03:15:25.000Z..2026-11-11T03:39:36.000Z.
Lab verification: run Unlighthouse on / with --mobile before and after the fix.
Target:
app/pages/_HomeDemoCarousel.vueVerify with:
pnpm dlx unlighthouse-ci --site http://localhost:4173 --urls / --mobile --throttle --samples 3 --reporter jsonExpanded --output-path "$HOME/scratch/vitals-review/hero-hydration/after"Estimated to change 1 file.