Releases: Krishna-Modi12/frontend-design-pro
Release list
v14.11.4
Release Notes — frontend-design-pro v14.11.4
Date: 2026-08-25
Archive: frontend-design-pro-v14.11.4.skill — 2136 KB, 418 files, ~535,779 tokens of markdown, root frontend-design-pro/
Pipeline wall-clock: 117.8s
Those archive figures describe the file this run produced, on the machine it ran on.
A local build can differ from the published one — line endings and any untracked working-tree
file move the byte count and the file count. The asset attached to the GitHub release is the
authoritative artifact; where the two disagree, believe the asset. Everything below this line
is recomputed from the repository and is identical either way.
Contents
19 skills · 8 core files · 111 references (402,592 tokens of on-demand depth) · 55 examples (45 gold + 10 anti-examples) · 45 tests · 61 constraints (17 semantic + 44 syntactic)
Registry (SKILL.md) is 2,112 tokens and is the only file always loaded.
Gate Results
| Gate | Result | Detail |
|---|---|---|
| Compile | PASS | All 55 gold examples compile under tsc --noEmit (strict) · 20 demo files clean |
| Semantic | PASS | 65/65 files (45 golds + 20 demo) pass 17/17 parser checks |
| Syntactic | PASS | gold examples clean, anti-examples fail as designed (44/44) · demos clean |
| References | PASS | 19 ban-shaped constraints over fenced code in every reference, skill and core file (0 violations) |
| Figures | PASS | 141 claim surfaces, figures derived from the filesystem (0 drifts) |
| Pipeline | PASS | 16/16 checks (stages · architecture · cited paths) |
| Evals | PASS | 22/22 self-test |
| Test coverage | PASS | 45/45 golds have a 1:1 test; all compile strict; 45/45 files · 232/232 tests pass |
| Regression | PASS | 14/14 synthetic cases |
| Figure patterns | PASS | 76 prose fixtures, both directions |
Plus pre-flight, frontmatter, path integrity, and per-skill budget gates — all blocking.
Known gaps
See ARCHITECTURE.md. Summary: the suite runs against
test/stubs/, not the examples' ~25 real peer libraries, so it proves the components
mount, expose the roles they claim and survive axe — not that they work against the
real three or react-hook-form; and reference depth is unevenly distributed across
skills.
Install
unzip frontend-design-pro-v14.11.4.skill -d ~/.claude/skills/See INSTALL.md and USAGE.md.
All gates passed. No manual changes were made after gate passage — the archive is a
deterministic product of a clean origin/main checkout (Stage 4.5 refuses to build from
anything else), re-verified against its own unzipped copy for both compilation and
content (Stage 6).
Released by: build_release.py
What's Changed
- home: restore hero particle-typography effect, fix shader visibility by @Krishna-Modi12 in #85
- feat(design-system): scene-scoped type baselines, and when a heading may be decorated by @Krishna-Modi12 in #86
- home: add a real favicon by @Krishna-Modi12 in #87
- feat: complete the xiaopu-ai extraction — six references, the token extractor, and the defects fixed on the way in by @Krishna-Modi12 in #88
- feat(extraction): xiaopu tooling, worked DESIGN.md example, and a tw-typography correction by @Krishna-Modi12 in #89
- docs: capability matrix, research log, and initiative record by @Krishna-Modi12 in #90
- fix(references): three wrong Radix claims, a routing gap, a grapheme bug, and the ingestion review by @Krishna-Modi12 in #91
- feat: visual regression baseline system by @Krishna-Modi12 in #92
- docs: WCAG 2.2 machine-check deferral, with reasoning by @Krishna-Modi12 in #93
- chore: re-baseline visual-regression from CI by @Krishna-Modi12 in #94
- feat: security scan for javascript: URLs, postMessage wildcards, raw innerHTML by @Krishna-Modi12 in #95
- i18n/RTL: layout-resilience check for home/ by @Krishna-Modi12 in #96
- Design-token intelligence report (Phase 2) by @Krishna-Modi12 in #97
- State-matrix reference: formalize loading/empty/error/success per surface (Phase 2) by @Krishna-Modi12 in #98
- SLOP-06: ban custom mouse cursors (Phase 3 — Anti-slop 2.0, scoped honestly) by @Krishna-Modi12 in #100
- fix(animations): one canonical reveal config; Lenis vs CSS scroll-behavior by @Krishna-Modi12 in #99
- Fix stale 60->61 constraint count in CAPABILITY_MATRIX.md by @Krishna-Modi12 in #101
Full Changelog: v14.11.3...v14.11.4
v14.11.3
Release Notes — frontend-design-pro v14.11.3
Date: 2026-08-22
Archive: frontend-design-pro-v14.11.3.skill — 2059 KB, 406 files, ~502,867 tokens of markdown, root frontend-design-pro/
Pipeline wall-clock: 107.2s
Those archive figures describe the file this run produced, on the machine it ran on.
A local build can differ from the published one — line endings and any untracked working-tree
file move the byte count and the file count. The asset attached to the GitHub release is the
authoritative artifact; where the two disagree, believe the asset. Everything below this line
is recomputed from the repository and is identical either way.
Contents
19 skills · 8 core files · 104 references (372,212 tokens of on-demand depth) · 55 examples (45 gold + 10 anti-examples) · 45 tests · 60 constraints (17 semantic + 43 syntactic)
Registry (SKILL.md) is 2,112 tokens and is the only file always loaded.
Gate Results
| Gate | Result | Detail |
|---|---|---|
| Compile | PASS | All 55 gold examples compile under tsc --noEmit (strict) · 20 demo files clean |
| Semantic | PASS | 65/65 files (45 golds + 20 demo) pass 17/17 parser checks |
| Syntactic | PASS | gold examples clean, anti-examples fail as designed (43/43) · demos clean |
| References | PASS | 19 ban-shaped constraints over fenced code in every reference, skill and core file (0 violations) |
| Figures | PASS | 131 claim surfaces, figures derived from the filesystem (0 drifts) |
| Pipeline | PASS | 16/16 checks (stages · architecture · cited paths) |
| Evals | PASS | 22/22 self-test |
| Test coverage | PASS | 45/45 golds have a 1:1 test; all compile strict; 45/45 files · 232/232 tests pass |
| Regression | PASS | 14/14 synthetic cases |
| Figure patterns | PASS | 76 prose fixtures, both directions |
Plus pre-flight, frontmatter, path integrity, and per-skill budget gates — all blocking.
Known gaps
See ARCHITECTURE.md. Summary: the suite runs against
test/stubs/, not the examples' ~25 real peer libraries, so it proves the components
mount, expose the roles they claim and survive axe — not that they work against the
real three or react-hook-form; and reference depth is unevenly distributed across
skills.
Install
unzip frontend-design-pro-v14.11.3.skill -d ~/.claude/skills/See INSTALL.md and USAGE.md.
All gates passed. No manual changes were made after gate passage — the archive is a
deterministic product of a clean origin/main checkout (Stage 4.5 refuses to build from
anything else), re-verified against its own unzipped copy for both compilation and
content (Stage 6).
Released by: build_release.py
What's Changed
- polish: pause marquee off-viewport, cinematic scroll-linked hero by @Krishna-Modi12 in #84
Full Changelog: v14.11.2...v14.11.3
v14.11.2
Release Notes — frontend-design-pro v14.11.2
Date: 2026-08-22
Archive: frontend-design-pro-v14.11.2.skill — 2058 KB, 406 files, ~502,404 tokens of markdown, root frontend-design-pro/
Pipeline wall-clock: 90.2s
Those archive figures describe the file this run produced, on the machine it ran on.
A local build can differ from the published one — line endings and any untracked working-tree
file move the byte count and the file count. The asset attached to the GitHub release is the
authoritative artifact; where the two disagree, believe the asset. Everything below this line
is recomputed from the repository and is identical either way.
Contents
19 skills · 8 core files · 104 references (372,212 tokens of on-demand depth) · 55 examples (45 gold + 10 anti-examples) · 45 tests · 60 constraints (17 semantic + 43 syntactic)
Registry (SKILL.md) is 2,112 tokens and is the only file always loaded.
Gate Results
| Gate | Result | Detail |
|---|---|---|
| Compile | PASS | All 55 gold examples compile under tsc --noEmit (strict) · 20 demo files clean |
| Semantic | PASS | 65/65 files (45 golds + 20 demo) pass 17/17 parser checks |
| Syntactic | PASS | gold examples clean, anti-examples fail as designed (43/43) · demos clean |
| References | PASS | 19 ban-shaped constraints over fenced code in every reference, skill and core file (0 violations) |
| Figures | PASS | 131 claim surfaces, figures derived from the filesystem (0 drifts) |
| Pipeline | PASS | 16/16 checks (stages · architecture · cited paths) |
| Evals | PASS | 22/22 self-test |
| Test coverage | PASS | 45/45 golds have a 1:1 test; all compile strict; 45/45 files · 232/232 tests pass |
| Regression | PASS | 14/14 synthetic cases |
| Figure patterns | PASS | 76 prose fixtures, both directions |
Plus pre-flight, frontmatter, path integrity, and per-skill budget gates — all blocking.
Known gaps
See ARCHITECTURE.md. Summary: the suite runs against
test/stubs/, not the examples' ~25 real peer libraries, so it proves the components
mount, expose the roles they claim and survive axe — not that they work against the
real three or react-hook-form; and reference depth is unevenly distributed across
skills.
Install
unzip frontend-design-pro-v14.11.2.skill -d ~/.claude/skills/See INSTALL.md and USAGE.md.
All gates passed. No manual changes were made after gate passage — the archive is a
deterministic product of a clean origin/main checkout (Stage 4.5 refuses to build from
anything else), re-verified against its own unzipped copy for both compilation and
content (Stage 6).
Released by: build_release.py
What's Changed
- home: reduce card copy to visual proof (Sections 02-04) by @Krishna-Modi12 in #83
Full Changelog: v14.11.1...v14.11.2
v14.11.1
Release Notes — frontend-design-pro v14.11.1
Date: 2026-08-22
Archive: frontend-design-pro-v14.11.1.skill — 2056 KB, 406 files, ~501,667 tokens of markdown, root frontend-design-pro/
Pipeline wall-clock: 121.7s
Those archive figures describe the file this run produced, on the machine it ran on.
A local build can differ from the published one — line endings and any untracked working-tree
file move the byte count and the file count. The asset attached to the GitHub release is the
authoritative artifact; where the two disagree, believe the asset. Everything below this line
is recomputed from the repository and is identical either way.
Contents
19 skills · 8 core files · 104 references (371,905 tokens of on-demand depth) · 55 examples (45 gold + 10 anti-examples) · 45 tests · 60 constraints (17 semantic + 43 syntactic)
Registry (SKILL.md) is 2,112 tokens and is the only file always loaded.
Gate Results
| Gate | Result | Detail |
|---|---|---|
| Compile | PASS | All 55 gold examples compile under tsc --noEmit (strict) · 20 demo files clean |
| Semantic | PASS | 65/65 files (45 golds + 20 demo) pass 17/17 parser checks |
| Syntactic | PASS | gold examples clean, anti-examples fail as designed (43/43) · demos clean |
| References | PASS | 19 ban-shaped constraints over fenced code in every reference, skill and core file (0 violations) |
| Figures | PASS | 129 claim surfaces, figures derived from the filesystem (0 drifts) |
| Pipeline | PASS | 16/16 checks (stages · architecture · cited paths) |
| Evals | PASS | 22/22 self-test |
| Test coverage | PASS | 45/45 golds have a 1:1 test; all compile strict; 45/45 files · 232/232 tests pass |
| Regression | PASS | 14/14 synthetic cases |
| Figure patterns | PASS | 76 prose fixtures, both directions |
Plus pre-flight, frontmatter, path integrity, and per-skill budget gates — all blocking.
Known gaps
See ARCHITECTURE.md. Summary: the suite runs against
test/stubs/, not the examples' ~25 real peer libraries, so it proves the components
mount, expose the roles they claim and survive axe — not that they work against the
real three or react-hook-form; and reference depth is unevenly distributed across
skills.
Install
unzip frontend-design-pro-v14.11.1.skill -d ~/.claude/skills/See INSTALL.md and USAGE.md.
All gates passed. No manual changes were made after gate passage — the archive is a
deterministic product of a clean origin/main checkout (Stage 4.5 refuses to build from
anything else), re-verified against its own unzipped copy for both compilation and
content (Stage 6).
Released by: build_release.py
What's Changed
- Rebuild the pack's own homepage as a real Next app (home/) by @Krishna-Modi12 in #81
- Homepage v2.1: prove instead of telling by @Krishna-Modi12 in #82
Full Changelog: v14.11.0...v14.11.1
v14.11.0
Release Notes — frontend-design-pro v14.11.0
Date: 2026-08-21
Archive: frontend-design-pro-v14.11.0.skill — 2069 KB, 406 files, ~511,075 tokens of markdown, root frontend-design-pro/
Pipeline wall-clock: 92.6s
Those archive figures describe the file this run produced, on the machine it ran on.
A local build can differ from the published one — line endings and any untracked working-tree
file move the byte count and the file count. The asset attached to the GitHub release is the
authoritative artifact; where the two disagree, believe the asset. Everything below this line
is recomputed from the repository and is identical either way.
Contents
19 skills · 8 core files · 104 references (371,905 tokens of on-demand depth) · 55 examples (45 gold + 10 anti-examples) · 45 tests · 60 constraints (17 semantic + 43 syntactic)
Registry (SKILL.md) is 2,099 tokens and is the only file always loaded.
Gate Results
| Gate | Result | Detail |
|---|---|---|
| Compile | PASS | All 55 gold examples compile under tsc --noEmit (strict) · 20 demo files clean |
| Semantic | PASS | 65/65 files (45 golds + 20 demo) pass 17/17 parser checks |
| Syntactic | PASS | gold examples clean, anti-examples fail as designed (43/43) · demos clean |
| References | PASS | 19 ban-shaped constraints over fenced code in every reference, skill and core file (0 violations) |
| Figures | PASS | 96 claim surfaces, figures derived from the filesystem (0 drifts) |
| Pipeline | PASS | 16/16 checks (stages · architecture · cited paths) |
| Evals | PASS | 22/22 self-test |
| Test coverage | PASS | 45/45 golds have a 1:1 test; all compile strict; 45/45 files · 232/232 tests pass |
| Regression | PASS | 14/14 synthetic cases |
| Figure patterns | PASS | 76 prose fixtures, both directions |
Plus pre-flight, frontmatter, path integrity, and per-skill budget gates — all blocking.
Known gaps
See ARCHITECTURE.md. Summary: the suite runs against
test/stubs/, not the examples' ~25 real peer libraries, so it proves the components
mount, expose the roles they claim and survive axe — not that they work against the
real three or react-hook-form; and reference depth is unevenly distributed across
skills.
Install
unzip frontend-design-pro-v14.11.0.skill -d ~/.claude/skills/See INSTALL.md and USAGE.md.
All gates passed. No manual changes were made after gate passage — the archive is a
deterministic product of a clean origin/main checkout (Stage 4.5 refuses to build from
anything else), re-verified against its own unzipped copy for both compilation and
content (Stage 6).
Released by: build_release.py
What's Changed
- docs: rebuild the landing page dark, tight and alive by @Krishna-Modi12 in #69
- fix: everything a cold install and one real page turned up by @Krishna-Modi12 in #70
- fix: the rules the golds break, and the ones no gate reads by @Krishna-Modi12 in #71
- docs: give the gallery an index and move the showcase off the root by @Krishna-Modi12 in #73
- docs: say what
npx skills addactually installs, and what the live demos actually track by @Krishna-Modi12 in #72 - fix(constraints): make the anti-slop wall enforce the half of itself nobody read by @Krishna-Modi12 in #74
- docs: make the README show the thing instead of describing it by @Krishna-Modi12 in #75
- docs: add the one section the reference README has and this one lacked by @Krishna-Modi12 in #76
- Rebuild the Pages site as a product page you can use by @Krishna-Modi12 in #79
- Take the reference sites' typographic scale by @Krishna-Modi12 in #80
- chore: conform to Anthropic's official skill-creator spec (and make the gates see it) by @Krishna-Modi12 in #78
- feat(references): typographic finishing, motion budget, Radix primitives — and the two figure shapes that hid eight stale claims by @Krishna-Modi12 in #77
Full Changelog: v14.10.1...v14.11.0
v14.10.1
Release Notes — frontend-design-pro v14.10.1
Date: 2026-08-14
Archive: frontend-design-pro-v14.10.1.skill — 2057 KB, 401 files, ~481,181 tokens of markdown, root frontend-design-pro/
Pipeline wall-clock: 119.6s
Contents
19 skills · 8 core files · 101 references (349,445 tokens of on-demand depth) · 55 examples (45 gold + 10 anti-examples) · 45 tests · 59 constraints (17 semantic + 42 syntactic)
Registry (SKILL.md) is 2,099 tokens and is the only file always loaded.
Gate Results
| Gate | Result | Detail |
|---|---|---|
| Compile | PASS | All 55 gold examples compile under tsc --noEmit (strict) · 19 demo files clean |
| Semantic | PASS | 64/64 files (45 golds + 19 demo) pass 17/17 parser checks |
| Syntactic | PASS | gold examples clean, anti-examples fail as designed (42/42) · demos clean |
| References | PASS | 19 ban-shaped constraints over fenced code in every reference, skill and core file (0 violations) |
| Figures | PASS | 89 claim surfaces, figures derived from the filesystem (0 drifts) |
| Pipeline | PASS | 16/16 checks (stages · architecture · cited paths) |
| Evals | PASS | 22/22 self-test |
| Test coverage | PASS | 45/45 golds have a 1:1 test; all compile strict; 45/45 files · 229/229 tests pass |
| Regression | PASS | 13/13 synthetic cases |
| Figure patterns | PASS | 20 prose fixtures, both directions |
Plus pre-flight, frontmatter, path integrity, and per-skill budget gates — all blocking.
Known gaps
See ARCHITECTURE.md. Summary: the suite runs against
test/stubs/, not the examples' ~25 real peer libraries, so it proves the components
mount, expose the roles they claim and survive axe — not that they work against the
real three or react-hook-form; and reference depth is unevenly distributed across
skills.
Install
unzip frontend-design-pro-v14.10.1.skill -d ~/.claude/skills/See INSTALL.md and USAGE.md.
All gates passed. No manual changes were made after gate passage — the archive is a
deterministic product of a clean origin/main checkout (Stage 4.5 refuses to build from
anything else), re-verified against its own unzipped copy for both compilation and
content (Stage 6).
Released by: build_release.py
What's Changed
- demo: align the landing page to one rail, and add an animated README hero by @Krishna-Modi12 in #64
- docs: rebuild the README hero as a component audit, and sharpen the headline by @Krishna-Modi12 in #65
- Remove landing page demo badge from README by @Krishna-Modi12 in #66
- fix(figures): teach Gate 11 the prose forms it could not read by @Krishna-Modi12 in #67
- fix(figures): gate the fixture count that went stale on its way out by @Krishna-Modi12 in #68
Full Changelog: v14.10.0...v14.10.1
v14.10.0
Release Notes — frontend-design-pro v14.10.0
Date: 2026-08-13
Archive: frontend-design-pro-v14.10.0.skill — 2050 KB, 399 files, ~478,684 tokens of markdown, root frontend-design-pro/
Pipeline wall-clock: 117.6s
Contents
19 skills · 8 core files · 101 references (349,445 tokens of on-demand depth) · 55 examples (45 gold + 10 anti-examples) · 45 tests · 59 constraints (17 semantic + 42 syntactic)
Registry (SKILL.md) is 2,099 tokens and is the only file always loaded.
Gate Results
| Gate | Result | Detail |
|---|---|---|
| Compile | PASS | All 55 gold examples compile under tsc --noEmit (strict) · 18 demo files clean |
| Semantic | PASS | 63/63 files (45 golds + 18 demo) pass 17/17 parser checks |
| Syntactic | PASS | gold examples clean, anti-examples fail as designed (42/42) · demos clean |
| References | PASS | 19 ban-shaped constraints over fenced code in every reference, skill and core file (0 violations) |
| Figures | PASS | 85 claim surfaces, figures derived from the filesystem (0 drifts) |
| Pipeline | PASS | 16/16 checks (stages · architecture · cited paths) |
| Evals | PASS | 22/22 self-test |
| Test coverage | PASS | 45/45 golds have a 1:1 test; all compile strict; 45/45 files · 229/229 tests pass |
| Regression | PASS | 13/13 synthetic cases |
Plus pre-flight, frontmatter, path integrity, and per-skill budget gates — all blocking.
Known gaps
See ARCHITECTURE.md. Summary: the suite runs against
test/stubs/, not the examples' ~25 real peer libraries, so it proves the components
mount, expose the roles they claim and survive axe — not that they work against the
real three or react-hook-form; and reference depth is unevenly distributed across
skills.
Install
unzip frontend-design-pro-v14.10.0.skill -d ~/.claude/skills/See INSTALL.md and USAGE.md.
All gates passed. No manual changes were made after gate passage — the archive is a
deterministic product of a clean origin/main checkout (Stage 4.5 refuses to build from
anything else), re-verified against its own unzipped copy for both compilation and
content (Stage 6).
Released by: build_release.py
What's Changed
- feat: say what changes when the pointer is precise by @Krishna-Modi12 in #52
- build: pre-flight catches a stale README release heading by @Krishna-Modi12 in #53
- fix: say where the 3D and scroll material actually came from by @Krishna-Modi12 in #54
- feat: write the two promised references that survive dedup by @Krishna-Modi12 in #55
- fix: correct the figures Gate 11 was never able to see by @Krishna-Modi12 in #56
- fix: match DESIGN.md headings on words, not on string equality by @Krishna-Modi12 in #57
- demo: rebuild the landing page as light editorial sample output by @Krishna-Modi12 in #58
- feat: the animation failures that survive a passing build by @Krishna-Modi12 in #59
- feat: the typography rules that invert for CJK by @Krishna-Modi12 in #60
- feat: say how to derive a palette, in the file that ships 161 of them by @Krishna-Modi12 in #61
- fix: the Aceternity domain was wrong, and the licence unstated by @Krishna-Modi12 in #62
- docs: release notes for the ingestion sprint by @Krishna-Modi12 in #63
Full Changelog: v14.9.0...v14.10.0
v14.9.0
Release Notes — frontend-design-pro v14.9.0
Date: 2026-08-12
Archive: frontend-design-pro-v14.9.0.skill — 1892 KB, 392 files, ~460,700 tokens of markdown, root frontend-design-pro/
Pipeline wall-clock: 108.6s
Contents
19 skills · 8 core files · 96 references (337,392 tokens of on-demand depth) · 55 examples (45 gold + 10 anti-examples) · 45 tests · 59 constraints (17 semantic + 42 syntactic)
Registry (SKILL.md) is 2,088 tokens and is the only file always loaded.
Gate Results
| Gate | Result | Detail |
|---|---|---|
| Compile | PASS | All 55 gold examples compile under tsc --noEmit (strict) · 16 demo files clean |
| Semantic | PASS | 61/61 files (45 golds + 16 demo) pass 17/17 parser checks |
| Syntactic | PASS | gold examples clean, anti-examples fail as designed (42/42) · demos clean |
| References | PASS | 19 ban-shaped constraints over fenced code in every reference, skill and core file (0 violations) |
| Figures | PASS | 83 claim surfaces, figures derived from the filesystem (0 drifts) |
| Pipeline | PASS | 16/16 checks (stages · architecture · cited paths) |
| Evals | PASS | 22/22 self-test |
| Test coverage | PASS | 45/45 golds have a 1:1 test; all compile strict; 45/45 files · 229/229 tests pass |
| Regression | PASS | 13/13 synthetic cases |
Plus pre-flight, frontmatter, path integrity, and per-skill budget gates — all blocking.
Known gaps
See ARCHITECTURE.md. Summary: the suite runs against
test/stubs/, not the examples' ~25 real peer libraries, so it proves the components
mount, expose the roles they claim and survive axe — not that they work against the
real three or react-hook-form; and reference depth is unevenly distributed across
skills.
Install
unzip frontend-design-pro-v14.9.0.skill -d ~/.claude/skills/See INSTALL.md and USAGE.md.
All gates passed. No manual changes were made after gate passage — the archive is a
deterministic product of a clean origin/main checkout (Stage 4.5 refuses to build from
anything else), re-verified against its own unzipped copy for both compilation and
content (Stage 6).
Released by: build_release.py
What's Changed
- Close the outsider-facing gaps: link integrity, disclosure, contribution, deps by @Krishna-Modi12 in #36
- Gate 11 could not read across a hard wrap — five stale figures were hiding there by @Krishna-Modi12 in #37
- fix: decouple the archive link check from its own fix; 24 smoke tests get a reason to fail by @Krishna-Modi12 in #38
- The figure gate could not see comments or pass ratios — and "205 of 229 tests" was live in four docs by @Krishna-Modi12 in #40
- feat: deploy the showcase to GitHub Pages by @Krishna-Modi12 in #42
- docs: link the live showcase from the README by @Krishna-Modi12 in #43
- Rebuild the README: header block, navigation, progressive disclosure by @Krishna-Modi12 in #41
- fix: the deployed showcase had no way home by @Krishna-Modi12 in #44
- feat: put the pack's own homepage at the front door by @Krishna-Modi12 in #45
- fix: the homepage published a gate count one behind its own source by @Krishna-Modi12 in #46
- Rebuild the landing page as a runnable anti-slop sample output by @Krishna-Modi12 in #47
- Link the landing page nobody could reach by @Krishna-Modi12 in #48
- Make every skill say when to load it, not just what it covers by @Krishna-Modi12 in #49
- release: v14.9.0 — the doctrine pointers, and the plugin manifest by @Krishna-Modi12 in #50
- docs: README announces the release it ships with by @Krishna-Modi12 in #51
Full Changelog: v14.8.1...v14.9.0
v14.8.1
Release Notes — frontend-design-pro v14.8.1
Date: 2026-08-11
Archive: frontend-design-pro-v14.8.1.skill — 1973 KB, 390 files, ~451,267 tokens of markdown, root frontend-design-pro/
Pipeline wall-clock: 109.7s
Contents
19 skills · 8 core files · 94 references (333,969 tokens of on-demand depth) · 55 examples (45 gold + 10 anti-examples) · 45 tests · 59 constraints (17 semantic + 42 syntactic)
Registry (SKILL.md) is 2,018 tokens and is the only file always loaded.
Gate Results
| Gate | Result | Detail |
|---|---|---|
| Compile | PASS | All 55 gold examples compile under tsc --noEmit (strict) · 17 demo files clean |
| Semantic | PASS | 62/62 files (45 golds + 17 demo) pass 17/17 parser checks |
| Syntactic | PASS | gold examples clean, anti-examples fail as designed (42/42) · demos clean |
| References | PASS | 19 ban-shaped constraints over fenced code in every reference, skill and core file (0 violations) |
| Figures | PASS | 74 claim surfaces, figures derived from the filesystem (0 drifts) |
| Pipeline | PASS | 16/16 checks (stages · architecture · cited paths) |
| Evals | PASS | 22/22 self-test |
| Test coverage | PASS | 45/45 golds have a 1:1 test; all compile strict; 45/45 files · 205/205 tests pass |
| Regression | PASS | 13/13 synthetic cases |
Plus pre-flight, frontmatter, path integrity, and per-skill budget gates — all blocking.
Known gaps
See ARCHITECTURE.md. Summary: the suite runs against
test/stubs/, not the examples' ~25 real peer libraries, so it proves the components
mount, expose the roles they claim and survive axe — not that they work against the
real three or react-hook-form; and reference depth is unevenly distributed across
skills.
Install
unzip frontend-design-pro-v14.8.1.skill -d ~/.claude/skills/See INSTALL.md and USAGE.md.
All gates passed. No manual changes were made after gate passage — the archive is a
deterministic product of a clean origin/main checkout (Stage 4.5 refuses to build from
anything else), re-verified against its own unzipped copy for both compilation and
content (Stage 6).
Released by: build_release.py
What's Changed
- fix: the archive shipped pointers to files it did not contain by @Krishna-Modi12 in #35
Full Changelog: v14.8.0...v14.8.1
v14.8.0
Release Notes — frontend-design-pro v14.8.0
Date: 2026-08-11
Archive: frontend-design-pro-v14.8.0.skill — 1927 KB, 377 files, ~427,591 tokens of markdown, root frontend-design-pro/
Pipeline wall-clock: 114.1s
Contents
19 skills · 8 core files · 94 references (333,969 tokens of on-demand depth) · 55 examples (45 gold + 10 anti-examples) · 45 tests · 59 constraints (17 semantic + 42 syntactic)
Registry (SKILL.md) is 2,018 tokens and is the only file always loaded.
Gate Results
| Gate | Result | Detail |
|---|---|---|
| Compile | PASS | All 55 gold examples compile under tsc --noEmit (strict) · 17 demo files clean |
| Semantic | PASS | 62/62 files (45 golds + 17 demo) pass 17/17 parser checks |
| Syntactic | PASS | gold examples clean, anti-examples fail as designed (42/42) · demos clean |
| References | PASS | 19 ban-shaped constraints over fenced code in every reference, skill and core file (0 violations) |
| Figures | PASS | 74 claim surfaces, figures derived from the filesystem (0 drifts) |
| Pipeline | PASS | 16/16 checks (stages · architecture · cited paths) |
| Evals | PASS | 22/22 self-test |
| Test coverage | PASS | 45/45 golds have a 1:1 test; all compile strict; 45/45 files · 205/205 tests pass |
| Regression | PASS | 13/13 synthetic cases |
Plus pre-flight, frontmatter, path integrity, and per-skill budget gates — all blocking.
Known gaps
See ARCHITECTURE.md. Summary: the suite runs against
test/stubs/, not the examples' ~25 real peer libraries, so it proves the components
mount, expose the roles they claim and survive axe — not that they work against the
real three or react-hook-form; and reference depth is unevenly distributed across
skills.
Install
unzip frontend-design-pro-v14.8.0.skill -d ~/.claude/skills/See INSTALL.md and USAGE.md.
All gates passed. No manual changes were made after gate passage — the archive is a
deterministic product of a clean origin/main checkout (Stage 4.5 refuses to build from
anything else), re-verified against its own unzipped copy for both compilation and
content (Stage 6).
Released by: build_release.py
What's Changed
- feat: Gate 11 — the figure gate by @Krishna-Modi12 in #34
Full Changelog: v14.7.4...v14.8.0