Repository navigation
Releases
v1.3.1 — Core Delivery Upgrade
🧰 Kio v1.3.1 — Core Delivery Upgrade
👑 kio-core-delivery 3.0.0
- New router (VI/EN): request triage (hotfix / bounded change / new feature), routing table, overlap rules, workflows and quality gates.
- Restored and added guides: Socratic brainstorming, Visual Companion, Spec-Kit SDD (
spec-kit.md,spec-driven.md), writing and executing plans, parallel subagent dispatch, Git worktrees, multi-CLI routing, 5-pillar executive audit, spec/plan reviewer prompts. - Templates: 5 Spec-Kit templates, 9 command workflows (
templates/commands/), 7 on-disk planning templates (task_plan.md,findings.md,progress.md…). - On-disk planning (planning-with-files): 20 scripts (
scripts/planning/: init-session, check-complete, plan-doctor, attest, ledger, inject…); paths adjusted for the Kio layout (kio-core-delivery/scripts/planning/); init-session, check-complete, plan-doctor and gate-stop smoke-tested. - Credits:
superpowers(Jesse Vincent), GitHub Spec-Kit,planning-with-files(Ahmad Adi), all MIT, in the skill's and the rootTHIRD_PARTY_NOTICES.md.
v1.3.0 — Automatic Stock Photos & Layout Checks for Frontend
🧰 Kio v1.3.0 — Automatic Stock Photos & Layout Checks for Frontend
🎨 kio-frontend-craft
- Free stock photos:
scripts/stock_images.pyfinds and fetches images from Unsplash / Pexels (with a key) or Openverse (no key), emits ready<img>tags withsrcset,width/height,alt, lazy loading and WebP variants, and writes credits toCREDITS.html. Each provider's rules are built in (Unsplash hotlink + download event, Pexels/Openverse download). Photos showing another brand are rejected on a brand's page. Live-tested through Openverse. - No HTML entities: write
&,©,–and accented letters directly;&, ,©,&#…;are banned, with agrepcheck before delivery. - No mobile edge spill: grids declare
grid-cols-1on mobile, child columns getmin-w-0, andoverflow-hiddenmay not hide a spill (found in the Ngo Gia test: a 368 px column in a 343 px gutter). - Rows stay aligned (rule 10): numbers above labels in stat rows, equal-height cards with price/CTA at the bottom, button labels never wrap,
tabular-numsfor figures. scripts/layout-check.js: console check at 375 px and 1440 px reportingoverflow,misalignedandwrappedButtons; validated on five pages with no false positives.- 150 tests pass (13 new for stock photos).
🧪 Real-world test on the home page
- The README gains a section comparing a seafood shop site built by Sonnet 5.5 without a skill and with
kio-frontend-craft: desktop/mobile screenshots, blind judge score (75 → 86/100), Lighthouse, SEO and layout checks, time, tokens and cost. Details and data inevaluations/showcase/seafood-v1.3/. - Lighthouse 100/100/100/100 as a rule: Google Fonts load non-blocking (preload + onload with a
noscriptfallback), a favicon is always declared, the hero image has correctsrcset/sizes. On the seafood page: Performance 98 → 100, Best Practices 96 → 100. layout-check.jsignores full-bleed background media (absoluteand as wide as its parent), previously reported as edge overflow.
🔧 Process
docs/RELEASE_WORKFLOW.mdadds a README and repository About review for every change;CLAUDE.mdrecords the convention for agents.
v1.2.0 — QA & Security Upgrade, Release-Pinned Installer, Complete Attribution
🧰 Kio v1.2.0 — QA & Security Upgrade, Release-Pinned Installer, Complete Attribution
🧪 kio-qa-security 3.0.0
-
Restored content lost in the skill consolidation: TDD and systematic debugging (root-cause tracing, defense in depth, condition-based waiting,
find-polluter.sh, TDD pressure scenarios), the 4 pstack deep-review modes (blast-radius, interrogate, arena, correct), the Playwright E2E guide, and the CodeQL/Semgrep workflows and scripts with their test suite. -
New:
references/codeql/SKILL.md(two-stage review, 5-layer checklist, "each Bash call is a fresh shell", CodeQL database discovery). 168/173 script tests pass; 2 more pass with Semgrep installed; 3 need an upstreamworkflows/codeql-build.jsthat was never shipped. -
New router (VI/EN), full English editions,
THIRD_PARTY_NOTICES.md. -
Answer language: the benchmark showed the Vietnamese
SKILL.mdmade the model answer an English prompt in Vietnamese (QA-01). The FE/BE/QA routers now tell the model to answer in the user's language (growth already did). Added after the benchmark run.
🛡️ Installer
curl | bash/irm | iexinstall the latest release tag instead ofmain; choose another version withKIO_REF.- Failed updates are no longer swallowed: the git error is printed, the summary says the old copy is still in use, and the installer exits
2; a failed fresh clone/download exits1. release.shrefuses to tag unless localmainequalsorigin/main.
🙏 Attribution
- 79 sub-skills record their original
author(Corey Haines, AgriciDaniel, puneetindersingh) withmaintainer: Kio System; the 5 SEOoriginal_authorcredits are restored. - 21 SEO
LICENSE.txtfiles carry the upstream copyright plus a Kio "Modifications" line. - The root
THIRD_PARTY_NOTICES.mdis now a complete index: adds Vercel, Anthropic, Profound, Knowatoa, ConversionKit, rowbound, Ploy, Phosphor, Google Fonts, pstack, Trail of Bits; inspirations: Ivan Falco, Gooseworks, spec-kit, Alibaba open-code-review.
🧪 Benchmarks
- Re-ran all 10 A/B scenarios on Claude Code (Sonnet 5.5 generation, blind Opus 5.5 judge, n = 1): average 7.61 → 8.01 (+0.40); 5 scenarios improved (FE-01 +3.8), 1 unchanged, 4 lower. The v1.0.2 results are kept as a historical section. Harness:
evaluations/harness/run_bench.py; evidence:evaluations/results-v1.2.0/.
v1.1.0 — Kio Skill Routers & Restored Content
🧰 Kio v1.1.0 — Skill Routers & Restored Content
📈 kio-growth-marketing 3.0.0
- New router over all 79 sub-skills (53 growth modules + 26 SEO Engine sub-skills), 17 merge rules for overlapping modules, 7 multi-module workflows.
- SEO Engine runs again: restored
seo-subskills/requirements.txt, remapped script paths, added run instructions. - 150 external links made relative, 25 broken links fixed; open-source credits consolidated in
THIRD_PARTY_NOTICES.md.
🎨 kio-frontend-craft 3.0.0
- Restored content lost in the 32→7 consolidation: 64 React/Next.js performance rules + full guide, UI/UX pro rules and quick reference, web performance, PDF → HTML, mobile (React Native/Expo, SwiftUI, offline).
- New: 9-pass
/unslopSOP and a Lenis smooth-scroll guide (checked against lenis 1.3.27). - New router with overlap rules and workflows. 137 tests pass.
⚙️ kio-backend-cloud 3.0.0
- Restored API & payments, database, DevOps, AI/ML, MCP builder and project scaffolding docs.
- New payments & idempotent webhooks guide (Stripe, PayPal, VNPay, MoMo, PayOS, SePay) with a tested reference implementation (
scripts/payment_signatures.py, 15 tests). - New MCP guides for the Python SDK 2.x (
MCPServer) and TypeScript SDK 1.32, with executed examples. - Expanded 9 thin references and fixed insecure/outdated snippets (RLS
WITH CHECK, pinned actions, slim-image healthchecks, Redis lock tokens, Terraform state locking). SQL verified on PostgreSQL 16.
🔧 Release tooling
- Includes v1.0.2 (piped installer auto-clone, A/B benchmark reports), now merged into
main. - English sync rewrites
*.en.mdlinks; machine-specificfile:///links replaced with relative paths.
v1.0.2 — Piped Auto-Clone Installer, A/B Benchmarks, Unified Versioning & Open-Source Attribution
Highlights (v1.0.2)
- Piped One-Line Installer Auto-Clone (
setup.sh&setup.ps1): Automatically clones/syncstoanpham2111/kio-agent-systemto~/.kio/repowhen executed viacurl ... | bashorirm ... | iex, with tarball/zip fallback. - Unified Version
v1.0.2: Synchronized version numbers across GitHub Releases, all 7SKILL.mdfrontmatters, and all benchmark logs. - Token-Optimized Skill Descriptions: Condensed YAML
descriptionfields to 1-2 high-signal trigger sentences and removed phantom sub-engine identifiers. - Internal Link Hygiene & MIT Attribution: Updated 150 internal references across
skills/to resolve insidetoanpham2111/kio-agent-systemand addedTHIRD_PARTY_NOTICES.md+ Credits section inREADME.md. - A/B Comparative Benchmark Overhaul: Upgraded all 10 scenarios in
evaluations/results/with side-by-side Baseline (Without Skill) vs. With Kio Skill raw code outputs and objective rubric scoring (+3.75 avg delta).
v1.0.1 — Production Release & Installer Security Patch
🧰 Kio System Release v1.0.1
Highlights
- 7 Modular Master Engineering Skills: Frontend Craft, Backend Cloud, QA Security, Core Delivery, Growth Marketing, WordPress FSE, Cognitive Thinking.
- Strict Installer Security & Ownership: Both
setup.sh(Linux/macOS) andsetup.ps1(Windows) verify realpath symlink targets and validate against~/.kio/manifest.json, preventing unauthorized overwrite or accidental deletion of personal symlinks. - Skill Evaluation Framework: 21 test scenarios defined with initial benchmark evidence logs across all 7 master skills in
evaluations/results/. - Community Governance & Standards: Full community documentation included (
CONTRIBUTING.md,CHANGELOG.md,SECURITY.md,CODE_OF_CONDUCT.md) under the MIT License.