Help us improve
Share bugs, ideas, or general feedback.
From sonzai-sdk
Use when the user is installing, configuring, or writing code against the Sonzai SDK (pip install sonzai, npm install @sonzai-labs/agents, go get github.com/sonz-ai/sonzai-go), calling api.sonz.ai, working with Sonzai agents/memory/personality/sessions, building a companion / matchmaker / personality-routed app / enterprise assistant / game NPC / coach / customer support agent, or migrating from raw HTTP curl calls to the typed SDK.
npx claudepluginhub sonz-ai/sonzai-claude-skill --plugin sonzai-sdkHow this skill is triggered — by the user, by Claude, or both
Slash command
/sonzai-sdk:sonzai-sdkThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Wizard-driven implementation guide for the Sonzai Mind Layer API across Python, TypeScript, and Go.
archetypes/coach-therapist.mdarchetypes/companion.mdarchetypes/customer-support.mdarchetypes/enterprise-assistant.mdarchetypes/game-npc.mdarchetypes/guide-router.mdarchetypes/hybrid-custom.mddecisions/byok-vs-customllm.mddecisions/capabilities-matrix.mddecisions/generation-vs-manual-create.mddecisions/instances-vs-multitenant.mddecisions/memory-mode.mddecisions/post-processing-model.mddecisions/proactive-channel.mddecisions/runtime-mode.mddecisions/sessions-vs-conversations.mddecisions/sharedmemory-vs-wisdom.mddecisions/state-vs-inventory.mdexisting-codebase-audit.mdfeatures/agent-insights.mdProvides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Provides UI/UX resources: 50+ styles, color palettes, font pairings, guidelines, charts for web/mobile across React, Next.js, Vue, Svelte, Tailwind, React Native, Flutter. Aids planning, building, reviewing interfaces.
Share bugs, ideas, or general feedback.
Wizard-driven implementation guide for the Sonzai Mind Layer API across Python, TypeScript, and Go.
sonzai, @sonzai-labs/agents, or github.com/sonz-ai/sonzai-gopip install sonzai, npm install @sonzai-labs/agents, go get github.com/sonz-ai/sonzai-goSONZAI_API_KEY in their env or .envapi.sonz.ai, mentions Sonzai agents, memory, personality, mood, sessionsfeatures/mcp-integration.md@sonzai-labs/openclaw-context) — load features/openclaw-integration.mdThe committed OpenAPI snapshot inside an installed SDK version can lag the live spec. Always run the drift check before writing non-trivial integration code.
→ Read references/drift-detection.md
The wizard diagnoses what you're building, prescribes the archetype + memory mode + capabilities, then drives you through spec → review → plan.
→ Read intake.md
| Operator has... | Wants... | Use |
|---|---|---|
| A scoping conversation with stakeholders | Interactive wizard → spec → plan (no build) | this skill (sonzai-sdk) |
| A meeting transcript + no human in loop | Unattended autonomous build → running app. Sends Slack/Gmail notification on completion (build done healthy / build done failing QA) when MCPs are enabled. | full-auto (sibling) |
| A meeting transcript + tech-lead in loop | Supervised build with 2 gates (masterplan + live-app review). Operator can reply to gates from Slack DM or Gmail (with MCPs enabled) — terminal isn't required. | cto-loop (sibling) |
Both full-auto and cto-loop deploy the built app locally with docker compose and run functional QA against the running stack. full-auto auto-fixes; cto-loop lets the tech-lead steer the fix loop via free-text feedback. Both share the same core machinery (tech-stack derivation/intake, brownfield audit, version-search hard rule, builder + fixer subagents). Both auto-dispatch a notifier subagent at terminal events when Gmail + Slack MCPs are enabled. cto-loop additionally polls for replies via Slack DM + Gmail unread (24h budget, paused-resumable).
Trigger phrases:
full-auto: "build from this transcript" / "go full auto" / "run full-auto" / "/full-auto"cto-loop: "build it and let me review" / "CTO in the loop" / "transcript with two gates"If the sonzai-internal-staff sibling skill is also installed in this plugin install, load it for additional workspace awareness when the operator is Sonzai staff. The skill is install-time gated by the sonz-ai plugin marketplace (category: internal), so external users never have it on disk. See that skill's own SKILL.md for what it adds.
If the user explicitly says "skip wizard" / "I know what I want" / they're mid-implementation and just need a syntax lookup:
| User task | Reference |
|---|---|
| Streaming chat (SSE) vs async polling | references/streaming-chat.md |
| Auth, env vars, base URL | references/auth-and-setup.md |
| Python syntax | references/python.md |
| TypeScript syntax | references/typescript.md |
| Go syntax | references/go.md |
| Raw HTTP → typed SDK | references/migration-from-http.md |
| Errors, 4xx/5xx | references/troubleshooting.md |
| Package version recommendations (any) | references/version-search.md |
SONZAI_API_KEY to a browser or mobile client. Server-side only.references/drift-detection.md.pip install sonzai==0.5.0, next@14, postgres:14, or any other version pulled from training memory. See references/version-search.md for what to verify and how.client.foo.bar() but my IDE says it doesn't exist" → drift. Step 0.references/migration-from-http.md).