AI Decision System
Replayable, offline-verifiable records for AI-assisted go/no-go decisions: experiment promotions, typed model answers and human-approved actions.
翻譯準備中
自動路由多代理管線、程式碼品質門檻 (CQ1‑CQ22)、測試品質門檻 (Q1‑Q17),以及捆綁的堆疊規則。這是您的 AI 程式碼助理缺少的品質層。
AI 翻譯
Auto-routing multi-agent pipelines, code quality gates (CQ1-CQ22), test quality gates (Q1-Q17), and bundled stack rules. The quality layer your AI coding assistant is missing.
開發工具Shell · PythonGitHub @greglas75上線中營運者未確認
試用過了嗎? 點一下就好,不需要登入。
自動啟動的多代理技能生態系統,支援 Claude Code、Codex 與 Cursor。包含 51 項技能、26 個代理、品質門檻、知識庫、對抗審查、內容寫作與優化。
AI 翻譯
Auto-activating, multi-agent skill ecosystem for Claude Code, Codex, and Cursor. 51 skills, 26 agents, quality gates, knowledge store, adversarial review, content writing & optimization.
取自儲存庫 README
正在載入完整 README。 以下為已儲存的摘錄。
Auto-activating, multi-agent skill ecosystem for Claude Code, Codex, Cursor, Antigravity, and Kimi Code.
58 skills, 50 sub-agent definitions (48 distinct roles), quality gates, knowledge store, session recovery, and structured workflows — all in one plugin.
Installs to Claude Code + Codex + Cursor + Antigravity + Kimi Code in one command. Restart your IDE after install.
Requires Claude Code 1.0.33+. Check with claude --version, update with claude update.
If skills don't appear after update (known Claude Code cache issue):
Then start a new session.
Codex / Cursor / Antigravity / Kimi Code檢視完整 README ↗Auto-activating, multi-agent skill ecosystem for Claude Code, Codex, Cursor, Antigravity, and Kimi Code.
58 skills, 50 sub-agent definitions (48 distinct roles), quality gates, knowledge store, session recovery, and structured workflows — all in one plugin.
curl -fsSL https://raw.githubusercontent.com/greglas75/zuvo/main/scripts/quick-install.sh | bash
Installs to Claude Code + Codex + Cursor + Antigravity + Kimi Code in one command. Restart your IDE after install.
Windows (Git Bash): Git Bash's curl uses schannel, which fails with CRYPT_E_NO_REVOCATION_CHECK (0x80092012) when Windows cannot reach the certificate's revocation server. Add --ssl-revoke-best-effort:
curl --ssl-revoke-best-effort -fsSL https://raw.githubusercontent.com/greglas75/zuvo/main/scripts/quick-install.sh | bash
Or skip curl entirely: git clone https://github.com/greglas75/zuvo.git && bash zuvo/scripts/quick-install.sh. The installer needs Python 3 (python3, python or py -3). The Microsoft Store python3 stub does not count; install Python from python.org or set ZUVO_PYTHON=<path>.
Requires Claude Code 1.0.33+. Check with
claude --version, update withclaude update.
# Add the Zuvo marketplace (one-time)
claude plugin marketplace add greglas75/zuvo-marketplace
# Install
claude plugin install zuvo
git clone https://github.com/greglas75/zuvo.git
cd zuvo
./scripts/install.sh codex
git clone https://github.com/greglas75/zuvo.git
cd zuvo
./scripts/install.sh cursor
git clone https://github.com/greglas75/zuvo.git
cd zuvo
./scripts/install.sh antigravity
./scripts/install.sh
claude plugin marketplace update zuvo-marketplace
claude plugin update zuvo@zuvo-marketplace
Or enable auto-updates:
/plugin → Select zuvo-marketplace → Enable auto-update
If skills don't appear after update (known Claude Code cache issue):
claude plugin uninstall zuvo@zuvo-marketplace
claude plugin install zuvo
Then start a new session.
cd /path/to/zuvo
git pull
./scripts/install.sh codex # or: cursor, antigravity, kimi, all
install.sh doesOne script, five platforms:
| Platform | What it does |
|---|---|
| Claude Code | Copies source files to plugin cache (~/.claude/plugins/cache/...) |
| Codex | Builds adapted distribution (path replacement, unicode normalization, TOML agent generation) then copies to ~/.codex/skills/ + ~/.codex/agents/ |
| Cursor | Builds adapted distribution (Cursor v3 agent frontmatter, flat agents with skill prefixes) then copies to ~/.cursor/skills/ + ~/.cursor/agents/ |
| Antigravity | Builds adapted distribution (build-antigravity-skills.sh) then copies to ~/.gemini/antigravity/skills/ + ~/.gemini/antigravity/shared/ |
| Kimi Code | Builds adapted distribution (build-kimi-skills.sh) then copies to ~/.kimi-code/skills/ + ~/.kimi-code/agents/, and merges [[hooks]] into ~/.kimi-code/config.toml. Keeps parallel sub-agents — Kimi has them |
./scripts/install.sh # all platforms (default)
./scripts/install.sh claude # Claude Code only
./scripts/install.sh codex # Codex only
./scripts/install.sh cursor # Cursor only
./scripts/install.sh antigravity # Antigravity only
./scripts/install.sh kimi # Kimi Code only
For testing changes without committing:
# Edit files in zuvo-plugin/, then:
./scripts/install.sh
# Restart Claude Code / Codex to pick up changes
Optional: add CodeSift MCP for deep code analysis:
# ~/.codex/config.toml
[mcp_servers.codesift]
command = "npx"
args = ["-y", "codesift-mcp"]
zuvo:brainstorm → zuvo:plan → zuvo:execute with multi-agent exploration, quality gates, and evidence-based reviewzuvo:ship → zuvo:deploy → zuvo:canary for the full post-code lifecycle.zuvo/context/. Resume after context compaction or crashes without losing progress| Platform | Status | Install |
|---|---|---|
| Claude Code | Stable | claude plugin install zuvo |
| Codex | Experimental | ./scripts/install.sh codex |
| Cursor | Stable (v3 sub-agents) | ./scripts/install.sh cursor |
| Category | Count | Skills |
|---|---|---|
| Pipeline | 5 | brainstorm, plan, execute, worktree, receive-review |
| Core | 5 | build, review, refactor, refactor-radar, debug |
| Code/Test audits | 5 | code-audit, test-audit, api-audit, security-audit, pentest |
| Infra audits | 7 | performance-audit, db-audit, dependency-audit, ci-audit, env-audit, infra-audit, container-audit |
| Structure/SEO/GEO | 6 | structure-audit, seo-audit, seo-fix, geo-audit, geo-fix, architecture |
| Content | 5 | content-audit, content-fix, content-migration, write-article, content-expand |
| Design | 3 | design, design-review, ui-design-team |
| Testing | 5 | write-tests, fix-tests, write-e2e, tests-performance, mutation-test |
| Accessibility | 1 | a11y-audit |
| Release | 5 | ship, deploy, canary, release-docs, retro |
| Utility | 10 | docs, presentation, backlog, incident, benchmark, agent-benchmark, using-zuvo, context-audit, skill-eval, profile-session |
| Lead Generation | 1 | leads |
./scripts/release.sh patch "fix: description"
./scripts/release.sh minor "feat: description"
This bumps version, commits, pushes, tags, updates marketplace SHA.
./scripts/install.sh
This syncs source to Claude Code cache + Codex + Cursor. No git push, no marketplace update.
package.json is forMetadata only (name, version, description). There are no npm dependencies. No npm install needed. The version field is read by release.sh and install.sh.
MIT