v2.11.0 — 2026年06月09日

Features

• Support root skill index entries — skill index now accepts entries at the root level of indexed repositories, allowing a single repo to contribute skills without nested directories (#304) — @luongnv89
• Add briiirussell cybersecurity-skills source to the curated skill index — 1 new skill source for cybersecurity-focused agent skills (#303) — @luongnv89
• Add badlogic/pi-skills source to the curated skill index — 1 new skill source from the pi-skills repository (#302) — @luongnv89

Chores

• Refresh indexed skill sources — re-ingested all enabled repos in data/skill-index-resources.json (#306) — @luongnv89

Full Changelog: v2.10.0...v2.11.0

Contributors

• @luongnv89

Assets 2

Loading

v2.10.0 — 2026年06月03日

Features

• New find-me-skills skill — a conversational skill for users who have a goal but don't know which Agent Skills they need. It collects intent, confirms the user's situation, queries the live asm catalog (asm search --available --json), dedupes and explains the matches, lays out a step-by-step path with per-step inputs/outputs, and on approval emits an installable BundleManifest the user runs with asm bundle install ./file.json. Uses the CLI as the runtime discovery surface, so it works on any machine — not just the asm repo (#281) — @luongnv89

Documentation

• Fix the stale landing docs page — correct the Node/npm prerequisite, add the pi provider (19), document asm disable/asm enable, and align on the canonical author field — @luongnv89
• Backfill v2.7.0–v2.9.0 in docs/CHANGELOG.md and refresh the README maintained line — @luongnv89

Chores

• Refresh indexed skill sources — re-ingested all enabled repos in data/skill-index-resources.json (31 updated, 0 unchanged, 0 failed, 1 skipped) (#299) — @luongnv89

Full Changelog: v2.9.0...v2.10.0

Contributors

• @luongnv89

Assets 2

Loading

v2.9.0 — 2026年05月29日

Features

• asm disable <target> / asm enable <target> — curate the active skill set without uninstalling, so agents stop accidentally invoking unintended or conflicting skills (e.g. two code-review skills, or the word "workflow" in conversation triggering a workflow skill). Disabling renames an instance's SKILL.md → SKILL.md.disabled so both asm's scanner and external agents stop discovering it; enabling reverses the rename. Targeting supports individual names, wildcards ('workflow*', '*-review'), prefixes (openspec:, asc-), bulk (--all), and --tool / --scope filters. State persists in ~/.config/agent-skill-manager/skill-state.json keyed by (skill ×ばつ tool ×ばつ scope). asm list reconstructs disabled instances and renders them dimmed with a [disabled] tag; JSON output carries a disabled flag. Symlinked multi-provider installs share one SKILL.md, so a toggle honestly affects every tool sharing that source and warns accordingly (#91) — @luongnv89

Documentation

• Bump landing-page skill count to 3,400+ in the page, og, and twitter meta descriptions (catalog now holds 3,493 skills) — @luongnv89

Chores

• Remove residual Bun references across the project — install.sh rewritten from Bun to npm (checks Node ≥ 18 / npm ≥ 9 and instructs on failure rather than auto-installing a runtime); docs/DEPLOYMENT.md, docs/DEVELOPMENT.md, CONTRIBUTING.md, prd.md, and the bug-report template corrected to Node/npm; skill/agent docs point bun run preindex → npm run preindex; vestigial Bun-absence test guards removed and argv fixtures switched from bun to node. The Bun.spawn/Bun.env patterns in src/security-auditor.ts and src/installer.ts are intentionally kept as detection vocabulary for auditing third-party skills. Also corrected stale @opentui/core references left by the ink migration and restored dropped eval/publish and bun:ffi build-leak test coverage (#293) — @luongnv89

Full Changelog: v2.8.0...v2.9.0

Contributors

• @luongnv89

Assets 2

Loading

v2.8.0 — 2026年05月20日

Features

• asm uninstall <name> — new command with full reverse-installation logic. Supports -s <scope> and -t, --tool <provider> filters, automatically relocates the real folder to another surviving provider when the canonical host is being removed, cleans up empty parent dirs, and shows scope/tool/relocation info in the confirmation prompt (#279) — @luongnv89
• New refresh-index skill — re-ingests every already-enabled repo in data/skill-index-resources.json in one guided workflow (sync, classify into updated/unchanged/failed/skipped, verify catalog, then a confirmation-gated commit + PR). Inverse of skill-index-updater (#290) — @luongnv89
• Reorder install tool list by user preference — DEFAULT_PROVIDERS now leads with claude, codex, opencode, pi, hermes, openclaw; added the pi provider; mergeWithDefaults now inserts newly-introduced default providers in their canonical priority slot (anchored to the nearest preceding default), so existing users see the new order on upgrade (#291) — @luongnv89
• Repo-derived bundles — repo indexing now emits zero/one/multiple bundle records per repository, with explicit bundle metadata merged with best-effort inference. Bundles are exposed via CLI bundle load/list and the website bundle catalog; provenance to source repo and skill paths is preserved. 196 inferred bundles refreshed across 28 repos (#275) — @luongnv89

Bug Fixes

• asm uninstall <name> -t <provider> now preserves the .skill-lock.json entry (source-tracking metadata: source, commitHash, ref) when other providers still have the skill installed, so subsequent asm list/asm update keep working on the survivors. When a real-folder relocation moves the canonical home to a kept provider the lock entry's provider field is repointed; when no relocation happened the entry is left intact. The full-uninstall path (no -t) still drops the entry as before (#284) — @luongnv89
• Uninstaller now cleans up partial state on EXDEV cp fallback failure — when cross-device move fails mid-copy, the partial destination is removed instead of being left as half-written content (#283) — @luongnv89
• Uninstaller surfaces relocation failures via non-zero exit code — relocation errors are no longer swallowed, so CI and scripted callers see the failure (#282) — @luongnv89
• fix(test): Bundle modify/export tests no longer leak installed skill copies into the user's real ~/.claude/skills/. Each test creates its source dir as <tmpRoot>/<frontmatter-name>/ so the installer's destination basename matches the cleanup lookup; a per-test assertSkillUninstalled guard fails the test if the install survives, plus a one-shot beforeAll sweep removes any cli-skill-for-* leftovers from prior runs (#288) — @luongnv89
• fix(ci): Replace apt-based trivy install with a binary download to avoid Azure mirror timeouts in the security workflow

Documentation

• Document asm bundle command and the predefined bundles in the README (#202) — @luongnv89

Chores

• Supply-chain hardening across npm and GitHub Actions — pinned versions, ignore-scripts hardening, lockfile checks, dependency cooldown declaration (#277) — @luongnv89
• Refresh indexed skill sources across all 30 curated repos — net skill count updated across the catalog (#295) — @luongnv89

Full Changelog: v2.7.0...v2.8.0

Contributors

• @luongnv89

Assets 2

Loading

Changelog

v2.7.0 — 2026年05月10日

Features

• Add Quick Start section to the catalog skill detail page — guided three-step workflow (Security Check, Quality Evaluation, Install) with copy-to-clipboard buttons (#273) — @luongnv89
• Add romainsimon/paperasse to the curated skill index — 6 French bureaucracy skills (commissaire-aux-comptes, comptable, controleur-fiscal, fiscaliste, notaire, syndic) covering administrative procedures and document handling (#267, #271) — @luongnv89
• Re-sync coreyhaines31/marketingskills — 1 new co-marketing skill added (#268) — @luongnv89

Bug Fixes

• fix(test): Update skill-detail tests for UI changes
• fix(test-spawn): Kill children on parent disconnect

Documentation

• docs(security): Skill audit for issue #269 — security scan of 284 installed skills against prompt injection and credential theft risks, zero strict-risk hits found (#270) — @luongnv89

Chores

• Re-sync all 30 curated skill sources — net skill count updated across the catalog
• Clean up repository and expand .gitignore — removed tracked backup files, added dev tool directories (.cursor, .aider, .continue, .windsurf, .zed, .fleet) and build/cache log patterns

Full Changelog: v2.6.2...v2.7.0

v2.6.2 — 2026年05月07日

Chores

• Index 2 new curated skill sources — warpdotdev/oz-skills (15 skills for Warp AI agents) and entireio/skills (5 cross-agent skills for context and session handoff); regenerated catalog totals 7036 skills across 31 repos (#264) — @luongnv89
• Dedupe same-name skills within a single repo by directory priority — when one repo ships a skill at multiple supported install paths, keep exactly one entry using a deterministic order (skills/ > .claude/skills/ > .agent/skills/ and .agents/skills/ > first occurrence); cross-repo duplicates remain independent and dropped paths are logged via --verbose (#265, #266) — @luongnv89

Full Changelog: v2.6.1...v2.6.2

v2.6.1 — 2026年05月03日

Bug Fixes

• asm link now handles non-existent paths and directory symlinks gracefully — bare registry-style names (e.g. asm link code-review) and missing paths exit 1 with a polished error and an asm install <name> suggestion instead of crashing with a Node stack trace; the top-level path probe now uses stat instead of lstat so directory symlinks are followed (re-linking an already-linked skill works), while inner-loop entry classification still uses lstat to avoid traversing nested symlinks (#262) — @luongnv89

Full Changelog: v2.6.0...v2.6.1

v2.6.0 — 2026年05月01日

Features

• Add skill-upstream-pr built-in skill — forks a target repo via gh, delegates the improvement loop to skill-auto-improver, and opens a friendly suggestion PR upstream with an asm eval before/after metrics table; mandatory preview/approval before any public action and a minimum-delta check skip trivial PRs (#244) — @luongnv89
• asm eval now accepts author as the canonical frontmatter field, with creator retained as a legacy alias — autofixer writes author: going forward, and legacy skills declaring creator: keep their score with no migration required (#243) — @luongnv89
• asm eval skill-best-practice provider (v1.0.0 → v1.1.0) — align with the latest skill-creator standard (v1.7.1): add xhigh to the effort enum, warn when descriptions exceed the 250-char runtime budget (truncation chops the negative-trigger clause), require metadata.version (semver-formatted), warn when metadata.author is missing, and require frontmatter name to match the parent directory (#246) — @luongnv89
• skill-auto-improver (v0.2.0 → v1.0.2) — adapt the skill itself and its workflow to the skill-creator standard, restructured around two gates: Gate 1 (skill-creator must-pass floor) and Gate 2 (asm-eval 85/8 quality floor); adds a Phase 1 frontmatter-normalization step after asm eval --fix, new references/skill-creator-checklist.md and references/frontmatter-audit.md playbooks, dual-gate report layout, and per-loop metadata.version bump for the target skill (#253) — @luongnv89
• Add Paramchoudhary/ResumeSkills as a curated source in the skill-index — 20 resume optimization and job-search skills (76 SKILL.md entries including cross-platform mirrors) covering ATS optimization, bullet writing, cover letters, LinkedIn profiles, interview prep, salary negotiation, and executive/academic/creative resumes (#256) — @luongnv89

Bug Fixes

• Website list now surfaces a muted relPath sub-label on rows whose owner/repo::name collides — plugin-bundle repos ship the same skill name at multiple install paths, and identical-looking cards (matching name, owner/repo, description, badges) made the duplicates indistinguishable; non-colliding rows are unchanged (#241) — @luongnv89
• asm install now treats skills/x-skill and ./skills/x-skill identically — when the input contains a path separator and resolves to an existing directory in the current working directory, it is treated as a local path rather than dispatched to the registry as a scoped name (#249) — @luongnv89
• asm install --path and --all now compose correctly — when --all is set and the resolved path contains skill subdirectories, ASM scans subdirectories instead of failing on missing SKILL.md; when a subpath is supplied with --all, duplicate detection is scoped to that prefix so unrelated dupes don't abort the install; the zero-skills error message is also scope-aware (#251, #252) — @luongnv89

Chores

• Refresh the bundled skills/skill-creator/ skill to the upstream version — pulls in updated references, scripts, and templates used by skill-auto-improver and skill-upstream-pr — @luongnv89
• Add a local-first security baseline plus CI mirror — pre-commit hook runs gitleaks (secrets), trivy (dependencies), and semgrep (static analysis) offline, exits non-zero on HIGH/CRITICAL, and writes JSON + Markdown reports under security/; .github/workflows/security.yml mirrors the same runner on push to main and pull requests, with the trivy DB cache keyed on a weekly stamp (#257) — @luongnv89
• Bring skill-index-updater up to the skill-creator standard — move README.md under docs/, normalize frontmatter (metadata.version + metadata.author, drop top-level version), rewrite the description under the 250-char runtime budget with a negative-trigger clause, and add When to Use / Example / Expected Output / Edge Cases sections; asm eval 76 (C) → 97 (A) with every category >= 8 (#258) — @luongnv89
• Re-sync all 27 enabled skill sources via bun run preindex — picks up upstream additions, removals, version/description bumps, and refreshed evalSummary + tokenCount for every skill; net delta +51 skills (6797 → 6848) (#260) — @luongnv89

Full Changelog: v2.5.0...v2.6.0

v2.5.0 — 2026年04月24日

Features

• Add a bundle builder UI to the catalog website — interactive cart flow lets users browse skills, add any to a persistent (localStorage) cart visible from the header across all routes, fill bundle metadata (name/description/author/tags), then either export a BundleManifest-conforming .json or publish a pre-filled feature request issue to promote the bundle. Validation mirrors the CLI's validateBundle() rules and truncation keeps the issue URL under GitHub's ~8KB pre-fill limit (#238, #240) — @luongnv89
• Add google/skills and antonbabenko/terraform-skill as curated sources in the skill-index — 13 Google product/technology skills plus a Terraform/OpenTofu skill for AI agents (#237) — @luongnv89
• asm eval now warns when README.md sits at the skill root — flags a common packaging mistake that previously slipped through scoring (#227, #234) — @luongnv89
• Redesign the website /changelog page with an editorial release-log aesthetic — two-column layout with a Fraunces + Instrument Serif masthead, a JetBrains Mono sticky version index, monospace metadata strip, outlined tag pills, and numbered receipt-style entries separated by dashed rules; self-scoped .cl-* styles preserve existing ENTRIES data verbatim and work in both light and dark themes — @luongnv89

Bug Fixes

• Deploy Skill Catalog workflow now also triggers on changes to website-src/** and package.json — previously release commits that bumped only those paths would not fire the workflow, leaving the live site stale (e.g. the v2.4.0 release left the nav pill showing 2.3.0 because skills.min.json was never regenerated on CI) — @luongnv89

Chores

• Remove the eu-project-ops predefined website bun...

Contributors

• @luongnv89
• @Mordris

Assets 2

Loading

v2.6.2 — 2026年05月07日

Chores

• Index 2 new curated skill sources — warpdotdev/oz-skills (15 skills for Warp AI agents) and entireio/skills (5 cross-agent skills for context and session handoff); regenerated catalog totals 7036 skills across 31 repos (#264) — @luongnv89
• Dedupe same-name skills within a single repo by directory priority — when one repo ships a skill at multiple supported install paths, keep exactly one entry using a deterministic order (skills/ > .claude/skills/ > .agent/skills/ and .agents/skills/ > first occurrence); cross-repo duplicates remain independent and dropped paths are logged via --verbose (#265, #266) — @luongnv89

Full Changelog: v2.6.1...v2.6.2

Contributors

• @luongnv89

Assets 2

Loading

v2.6.1 — 2026年05月03日

Bug Fixes

• asm link now handles non-existent paths and directory symlinks gracefully — bare registry-style names (e.g. asm link code-review) and missing paths exit 1 with a polished error and an asm install <name> suggestion instead of crashing with a Node stack trace; the top-level path probe now uses stat instead of lstat so directory symlinks are followed (re-linking an already-linked skill works), while inner-loop entry classification still uses lstat to avoid traversing nested symlinks (#262) — @luongnv89

Full Changelog: v2.6.0...v2.6.1

Contributors

• @luongnv89

Assets 2

Loading

v2.6.0 — 2026年05月01日

Features

• Add skill-upstream-pr built-in skill — forks a target repo via gh, delegates the improvement loop to skill-auto-improver, and opens a friendly suggestion PR upstream with an asm eval before/after metrics table; mandatory preview/approval before any public action and a minimum-delta check skip trivial PRs (#244) — @luongnv89
• asm eval now accepts author as the canonical frontmatter field, with creator retained as a legacy alias — autofixer writes author: going forward, and legacy skills declaring creator: keep their score with no migration required (#243) — @luongnv89
• asm eval skill-best-practice provider (v1.0.0 → v1.1.0) — align with the latest skill-creator standard (v1.7.1): add xhigh to the effort enum, warn when descriptions exceed the 250-char runtime budget (truncation chops the negative-trigger clause), require metadata.version (semver-formatted), warn when metadata.author is missing, and require frontmatter name to match the parent directory (#246) — @luongnv89
• skill-auto-improver (v0.2.0 → v1.0.2) — adapt the skill itself and its workflow to the skill-creator standard, restructured around two gates: Gate 1 (skill-creator must-pass floor) and Gate 2 (asm-eval 85/8 quality floor); adds a Phase 1 frontmatter-normalization step after asm eval --fix, new references/skill-creator-checklist.md and references/frontmatter-audit.md playbooks, dual-gate report layout, and per-loop metadata.version bump for the target skill (#253) — @luongnv89
• Add Paramchoudhary/ResumeSkills as a curated source in the skill-index — 20 resume optimization and job-search skills (76 SKILL.md entries including cross-platform mirrors) covering ATS optimization, bullet writing, cover letters, LinkedIn profiles, interview prep, salary negotiation, and executive/academic/creative resumes (#256) — @luongnv89

Bug Fixes

• Website list now surfaces a muted relPath sub-label on rows whose owner/repo::name collides — plugin-bundle repos ship the same skill name at multiple install paths, and identical-looking cards (matching name, owner/repo, description, badges) made the duplicates indistinguishable; non-colliding rows are unchanged (#241) — @luongnv89
• asm install now treats skills/x-skill and ./skills/x-skill identically — when the input contains a path separator and resolves to an existing directory in the current working directory, it is treated as a local path rather than dispatched to the registry as a scoped name (#249) — @luongnv89
• asm install --path and --all now compose correctly — when --all is set and the resolved path contains skill subdirectories, ASM scans subdirectories instead of failing on missing SKILL.md; when a subpath is supplied with --all, duplicate detection is scoped to that prefix so unrelated dupes don't abort the install; the zero-skills error message is also scope-aware (#251, #252) — @luongnv89

Chores

• Refresh the bundled skills/skill-creator/ skill to the upstream version — pulls in updated references, scripts, and templates used by skill-auto-improver and skill-upstream-pr — @luongnv89
• Add a local-first security baseline plus CI mirror — pre-commit hook runs gitleaks (secrets), trivy (dependencies), and semgrep (static analysis) offline, exits non-zero on HIGH/CRITICAL, and writes JSON + Markdown reports under security/; .github/workflows/security.yml mirrors the same runner on push to main and pull requests, with the trivy DB cache keyed on a weekly stamp (#257) — @luongnv89
• Bring skill-index-updater up to the skill-creator standard — move README.md under docs/, normalize frontmatter (metadata.version + metadata.author, drop top-level version), rewrite the description under the 250-char runtime budget with a negative-trigger clause, and add When to Use / Example / Expected Output / Edge Cases sections; asm eval 76 (C) → 97 (A) with every category >= 8 (#258) — @luongnv89
• Re-sync all 27 enabled skill sources via bun run preindex — picks up upstream additions, removals, version/description bumps, and refreshed evalSummary + tokenCount for every skill; net delta +51 skills (6797 → 6848) (#260) — @luongnv89

Full Changelog: v2.5.0...v2.6.0

Contributors

• @luongnv89

Assets 2

Loading

v2.5.0 — 2026年04月24日

Features

• Add a bundle builder UI to the catalog website — interactive cart flow lets users browse skills, add any to a persistent (localStorage) cart visible from the header across all routes, fill bundle metadata (name/description/author/tags), then either export a BundleManifest-conforming .json or publish a pre-filled feature request issue to promote the bundle. Validation mirrors the CLI's validateBundle() rules and truncation keeps the issue URL under GitHub's ~8KB pre-fill limit (#238, #240) — @luongnv89
• Add google/skills and antonbabenko/terraform-skill as curated sources in the skill-index — 13 Google product/technology skills plus a Terraform/OpenTofu skill for AI agents (#237) — @luongnv89
• asm eval now warns when README.md sits at the skill root — flags a common packaging mistake that previously slipped through scoring (#227, #234) — @luongnv89
• Redesign the website /changelog page with an editorial release-log aesthetic — two-column layout with a Fraunces + Instrument Serif masthead, a JetBrains Mono sticky version index, monospace metadata strip, outlined tag pills, and numbered receipt-style entries separated by dashed rules; self-scoped .cl-* styles preserve existing ENTRIES data verbatim and work in both light and dark themes — @luongnv89

Bug Fixes

• Deploy Skill Catalog workflow now also triggers on changes to website-src/** and package.json — previously release commits that bumped only those paths would not fire the workflow, leaving the live site stale (e.g. the v2.4.0 release left the nav pill showing 2.3.0 because skills.min.json was never regenerated on CI) — @luongnv89

Chores

• Remove the eu-project-ops predefined website bundle and update bundle tests and the website changelog bundle count to match the reduced set — @luongnv89

Full Changelog: v2.4.0...v2.5.0

Contributors

• @luongnv89

Assets 2

Loading

v2.4.0 — 2026年04月23日

Features

• Drop Bun from the toolchain — replace @opentui/core with ink for the TUI so CLI and TUI both run on Node >=18 alone, removing the bun:ffi native dependency (#224, #226) — @luongnv89
• Make bun optional at install time — CLI runs on Node, Bun is only required for interactive TUI mode (#221, #223) — @luongnv89
• Add skill-auto-improver built-in skill — eval-driven improvement loop that runs asm eval, applies deterministic --fix, then iterates per-category playbook edits until a skill clears the 85/8 quality floor (overallScore > 85 AND every category >= 8) or stops with a blocker report. Self-dogfooded: scores 100/100 with every category at 10/10 (#209, #218) — @luongnv89
• Ship 5 curated pre-defined bundles with ASM for popular workflows (frontend-dev, devops, ios-release, content-writing, eu-project-ops) — add --predefined flag on asm bundle list to show shipped bundles (#206, #211) — @luongnv89
• Add asm bundle modify and asm bundle export subcommands — non-interactive editing via --add, --remove, --description, --author, --tags, plus an interactive prompt mode and JSON export with --force / --json (#204, #205, #208) — @luongnv89
• Rewrite the ASM catalog website on React + Vite + Tailwind + shadcn/ui — replaces the 4.4k-line single-file index.html, retains full feature parity (catalog list, category tabs, facets, repo/sort selects, MiniSearch prefix + fuzzy search, pagination, bundle list, skill detail), and introduces a NeuronWiz-style sidebar + detail two-pane layout for catalog and bundles (#228, #229, #230, #231) — @luongnv89
• Add /bundles page to the ASM catalog website — each bundle card shows name, description, tags, included skills, and a copy-paste install command; deploy pipeline triggers on data/bundles/** changes (#207, #215) — @luongnv89
• Add /docs and /changelog SPA pages to the React website so header nav stays internal instead of bouncing to GitHub — @luongnv89
• Consolidate mobile header actions (theme toggle, GitHub link, nav items) into a single burger menu at <=768px with aria-expanded, aria-controls, Escape-to-close, and outside-click dismissal (#216, #217) — @luongnv89
• Pin skill-auto-improver to the top of the catalog as a featured skill, with a distinct visual treatment (#219) — @luongnv89
• Restore full website nav (Docs, Changelog, version pill, GitHub star count) and switch the catalog sidebar to react-window virtualization — cuts click latency on the 6,783-skill catalog from ~10s to instant (#232) — @luongnv89
• Use the real ASM nexus logo in the website header — @luongnv89

Performance

• Split catalog.json into a compact list + a MiniSearch index — faster initial load and smaller per-request payloads on the website (#214, #220) — @luongnv89

Bug Fixes

• Include relPath in the catalog dedup key so plugin-bundle variants aren't dropped — recovers ~3k installable targets that the owner/repo::name key had silently collapsed; adds regression tests asserting totalSkills === skills.length and unique installUrl per entry (#201, #203) — @luongnv89
• Opt into React Router v7 future flags (v7_startTransition, v7_relativeSplatPath) to silence v6 deprecation warnings — @luongnv89
• Import test accounts for missing providers so the test suite runs cleanly when optional providers aren't installed — @luongnv89

Changed

• Rename eval provider ID skill-creator → skill-best-practice to avoid collision with the real Anthropic skill-creator skill; include warning-severity checks in the score denominator; add per-check √/×ばつ/⚠ breakdown under each extra provider's score line — @luongnv89

Testing

• Isolate publisher and import tests from host gh CLI and skill-provider state via a fakeGhCli() helper — removes five flaky tests that depended on the developer's local environment (#167, #200) — @luongnv89
• Deflake runner timing assertions by bumping sleeps to 20ms and assertions to >=10ms for headroom against Math.round jitter — @luongnv89
• Stub ResizeObserver in jsdom smoke tests so react-window v2 renders under Vitest — @luongnv89

Chores

• Ignore .agents/ directory in git — @luongnv89

Full Changelog: v2.3.0...v2.4.0

Contributors

• @luongnv89

Assets 2

Loading