cardinalconseils/claude-starter - Page 2

positioning

Brand positioning, competitive framing, mission/vision, community strategy — April Dunford Obviously Awesome methodology plus Seth Godin brand philosophy frameworks

video-ai-direction

AI video generation direction for Kling API — prompt engineering, motion design, platform specs, text-to-video and image-to-video workflows for marketing content

agentic-os-builder

Agentic OS — three-layer architecture for turning any project into a structured, AI-navigable operating system. Use when scaffolding project domains, memory layers, or observability dashboards. Triggered by /cks:agentic-os, or when the user wants to organize…

code-excellence

Positive model of what good code looks like across five structural dimensions: naming, function boundaries, error handling, testability, and idiom conformance. Use when writing new code, reviewing quality, refactoring for clarity, or when a reviewer flags…

channel-brain

Wires inbound Claude Code channel events (Telegram, iMessage, fakechat) into the concierge brain — routes each message through Converse/Dispatch/Clarify, keys per-user memory, and adapts confirmation and permission behaviour for unattended always-on operation

honcho-memory

Optional self-hosted Honcho memory layer that AUGMENTS CKS file memory — adds theory-of-mind user representations (the dialectic) on top of the durable local files, keyed to the trusted CKS_ACTIVE_USER as the Honcho peer. File memory stays the…

user-memory

Durable cross-project per-user memory for the conversational agent — remembers preferences, history, and facts about each user across sessions and projects, keyed per channel sender with isolation

channel-setup

Provisioning knowledge for per-project Telegram (and other channel) agents — gives each CKS project its own bot, its own isolated config dir, and its own always-on process, so one host runs many project bots without collision. Distinct from channel-brain…

proactive-brain

Lets the always-on conversational agent initiate messages — on a scheduled wake it scans the active user's blockers, due reminders, and stale clarifications, then pushes a short nudge out through the channel reply tool instead of waiting to be asked

ai-agent-projects

AI agent project scaffolding templates for vibe coders. Knows how to scaffold voice agents (Telnyx + ElevenLabs), chat agents (OpenRouter + Supabase), multi-agent systems, RAG pipelines, and MCP servers. Use when /cks:kickstart detects an AI agent project…

control-plane

CKS v6 Control Plane — multi-agent orchestration system with personas, memory, RAID logs, and observability. Not a user-facing skill.

experts-core-expert-builder

Builder expert — pragmatic architecture, implementation, deployment. Pattern for experts synthesizing Jensen Huang, Guillermo Rauch, Kelsey Hightower.

experts-core-expert-debugger

Debugger expert — systematic root cause analysis, testing strategy, performance. Pattern for experts synthesizing Kent Beck, John Carmack, DJ Patil.

experts-core-expert-product

Product expert — user-centered features, UX, prioritization, metrics. Pattern for experts synthesizing Julie Zhuo, Jony Ive, Dieter Rams.

experts

Multi-expert persona system — builder, product, debugger, 22 specialists. Invoked by /cks:expert or auto-surfaced by lifecycle phases.

guardrails

Domain guardrail rules — generated at bootstrap time based on detected stack. Produces scoped .claude/rules/ files for security, testing, database, and docs.

hermes-agent

Hermes Agent integration for CKS — project scaffolding, AGENTS.md generation, skill mapping, and Hermes-native templates. Use when the user asks for Hermes support, multi-harness scaffolding, or when generating AGENTS.md. Maps Hermes concepts (skills, memory,…

mcp-server-authoring

Scaffold, build, and deploy MCP (Model Context Protocol) servers in TypeScript. MCP is the modern pattern for extending AI tools — your server exposes tools that Claude Code, Hermes, and other MCP clients can call. Use when the project needs custom AI tool…

schema-markup

When the user wants to add, fix, or optimize schema markup and structured data on their site. Also use when the user mentions 'schema markup,' 'structured data,' 'JSON-LD,' 'rich snippets,' or 'schema.org.'

conversation-state

Persists the live conversation thread per user so an always-on session resumes mid-thread after a restart — running summary, recent turns, active work, and any pending clarification awaiting the user's next reply

architecture

Architecture design and communication for target projects. Use when: producing a project-level ARCHITECTURE.md, capturing Architecture Decision Records (ADRs), generating Mermaid topology diagrams, or helping users communicate architectural choices to their…

concierge

Intent routing — maps natural language to CKS workflows, conversational project orchestrator, voice and Slack command routing

decision-memo

When the user wants to write an executive memo, build a persuasive narrative, present a recommendation to a senior stakeholder, or structure an argument using the Pyramid Principle. Use when PMC says "write the narrative for X", "how do I sell this…

kpi-architect

When the user wants to define the right KPIs, stress-test a strategy through adversarial scenarios, identify and mitigate risks, or establish governance triggers. Use when PMC says "war game this", "what could kill this", "stress test", "what are the risks",…

market-mapping

When the user wants to map the competitive landscape, understand where profit concentrates, identify white space, or size the opportunity. Use when PMC says "map the market for X", "who are the real competitors", "who should I be targeting", "where's the…

mckinsey-strategy-os

McKinsey-style strategy OS — routes strategic analysis requests to the right module skill. Activate when PMC asks for strategic analysis, consulting-style frameworks, or structured business thinking. Covers: situation diagnosis, market mapping, strategic…

operating-model

When the user wants to design how the business actually runs, prioritize initiatives, or build a phased execution plan. Use when PMC says "how do I actually run this", "what should I build first", "operating model for X", "prioritize my initiatives",…

situation-assessment

When the user wants to diagnose a business situation before deciding what to do. Use when PMC says "what's actually going on with X", "help me frame this problem", "why isn't X working", "situation assessment", "what are the growth barriers", "am I making…

strategic-options

When the user wants to evaluate strategic options, develop a WHERE TO PLAY / HOW TO WIN framework, build a business case, or decide between two paths. Use when PMC says "what are my strategic options", "should I do X or Y", "build me a business case",…

grill-me

Relentless plan/design interrogator — one question at a time, recommends an answer per question, explores codebase before asking. Use when user says "grill me", "stress-test my plan", "interrogate this design", or invokes /cks:grill.

cks-canary

Post-deploy canary verification domain knowledge — console errors, page load, pass/fail criteria

scale-ladder

Progressive scaling decision tree for CKS projects. Encodes the 7-rung scaling ladder with maturity-gate sequencing and "not yet" guardrails. Delegates to monitoring, caching, performance, observability, and architecture skills — never duplicates their…

uat

End-of-feature User Acceptance Testing methodology — acceptance-criteria-driven, preflight-aware UAT using browser automation. Use when running /cks:uat, operating in BrowserUAT pipeline node, or verifying a feature's definition of done via browser.

design-fluency

Design vocabulary and visual-slop detection for UI output. Use when: generating screens, reviewing design adherence, running visual polish passes, detecting AI-generated visual clichés (overused fonts, gradient text, nested cards, monotonous spacing,…

evals

LLM output quality evaluation — memory/RAG evals, API response quality, tool-use correctness, prompt regression, safety/guardrails, and structured output validation. Use when: building any LLM feature, verifying AI output quality before shipping, running…

campaign

Campaign orchestration knowledge — campaign types, intake workflow, specialist dispatch map, Apollo integration decision tree, output schema for outbound, launch, ABM, and content/paid campaigns

debug

Debug anything — trace app runtime errors or diagnose CKS skill/agent/command issues. Language-agnostic diagnostic approach using code tracing and strategic logging. Use when: "debug", "why is this broken", "trace this error", "why did CKS do that", "what…

github-issues

Auto-file GitHub issues at CKS lifecycle events (verification failure, retro, backlog punt) and gate deploy/release on open blocking issues. Dark Factory: file then notify, never prompt before filing.

retrospective

Post-ship retrospective — extract conventions, patterns, velocity metrics, gotchas, CLAUDE.md improvements.

voice

Voice agent scaffolding — Telnyx AI Assistant + Call Control via MCP, Cloudflare Worker webhook, phone number provisioning directly from Claude