All Pages

TitleSectionDateDescription
tufte-viz — Core PrinciplesCompiled"2026-05-25"The core Tufte visualization principles encoded in the tufte-viz Claude Code skill — from the four Tufte books (Visual Display, Envisioning Information, Visual Explanations, Beautiful Evidence). Covers data-ink ratio, chartjunk, small multiples, lie factor, sparklines, layering, and the analytical design principles.
Vibe CodingCompiled2026-06-10Building applications rapidly by describing intent to an AI coding agent (Cursor, Codex, Claude Code), without writing code manually. The key shift: from writing code to reviewing and directing AI-generated code.
tufte-viz Skill — Origin StoryCompiled"2026-05-25"The origin story of Angelica Parente's tufte-viz Claude Code skill — a Tufte-principles visualization tool that went viral on May 24, 2026 after she shared it on X (Twitter). Originally created late January 2026 in a late-night coding session.
7-Agent Crew Topology — The Enterprise Crew ArchitectureCompiled2026-04-28The Enterprise Crew is the SuperAda multi-agent system: seven active AI agents spread across cloud VMs, a Raspberry Pi, and Mac hardware, all orchestrated via OpenClaw. Each agent has a defined role, …
SaaS is Dead (Vibe Coding Thesis)Compiled2026-06-10The thesis that generic, off-the-shelf SaaS is losing its value proposition as vibe coding enables anyone to build exactly what they need in a day — making horizontal SaaS tools vulnerable to custom-built alternatives.
Distribution GapCompiled"2026-06-09"The systematic failure to acquire traffic and revenue despite having production capability — the core problem underlying the factory trap.
Meta-Crons — The Four Agents That Keep the Fleet AliveCompiled2026-04-28Meta-crons are cron jobs that manage other cron jobs. When operating at fleet scale (~98 active autonomous tasks across the Enterprise Crew), ordinary monitoring and recovery is insufficient — broken …
Autonomy Policy v3 — Delegation That Actually WorksCompiled2026-04-28The Autonomy Policy v3 is the third and current iteration of the Enterprise Crew's operating policy for agent decision-making. It evolved from two earlier versions that each fixed a specific failure m…
Useless AI SlopCompiled"2026-06-09"Low-value AI-generated products that flood the market — the output of factory-trapped builders who optimize production over distribution.
Lobster Pipelines — Typed JSON Envelope with Resumable ApprovalsCompiled2026-04-28Lobster is the Enterprise Crew's workflow runtime for complex multi-step autonomous tasks. It consists of:
HiM Model — Human-in-the-Loop OrchestrationCompiled2026-04-28HiM stands for Human-in-the-Middle (or "Human is the Model" depending on framing). It is the foundational cognitive architecture of the SuperAda Enterprise Crew: a human provides vision, context, and …
Sharpener.devCompiled"2026-05-22"AI-powered product scoping tool that generates complete PRDs from a single-sentence idea. Built on the BMAD Method framework, it walks users through 9 structured sections: Discovery, Success Criteria, User Journeys, Domain, Innovation, Product Type, Scope, Functional Requirements, Non-Functional Requirements.
BMAD MethodCompiled"2026-05-22"Product management framework with 9 structured sections for turning rough ideas into complete PRDs — Discovery, Success Criteria, User Journeys, Domain, Innovation, Product Type, Scope, Functional Requirements, Non-Functional Requirements. Provides 50+ elicitation techniques per section for deeper pressure-testing.
World Model — Shared Cognitive Architecture for Multi-Agent SystemsCompiled2026-04-28The World Model is the Enterprise Crew's shared cognitive substrate: a single world.json file that serves as the authoritative source of truth for agent state, context, and coordination. It is the ans…
tufte-viz — Pre/Post Demo ReferenceCompiled"2026-05-25"Four worked pre/post demos showing how the tufte-viz skill transforms default visualizations into Tufte-principled ones. Each demo pairs a "before" (default output) with an "after" (Tufte-treatment) side by side on a single page.
Factory TrapCompiled"2026-06-09"The systematic pattern where builders over-invest in factory/agent tooling and under-invest in marketing/distribution — resulting in sophisticated pipelines that produce nothing people want.
Internal Tool Custom FitCompiled2026-06-10The advantage of building tools exactly to your workflow (custom fit) versus adapting your workflow to off-the-shelf SaaS. The key insight: vibe coding makes custom fit free, collapsing the traditional trade-off between 'build custom' and 'buy generic'.
ISC — Inference Selection CriteriaCompiled2026-04-28Inference Selection Criteria (ISC) are the standard for writing task criteria in the SuperAda Enterprise Crew's formal execution algorithm. They define what makes a task criterion well-formed enough t…
Kelly Tweets — Design, UX & DevelopmentCompiled2026-04-27Design Station: One prompt → 10 design agents crowd-sourcing trends from X, generating specs with Gemini, auto-applying styles. One of Kelly's most significant factory upgrades.
Chapter 10: Browser Agent Deep DiveCompiled2026-04-27The browser tool is OpenClaw's heaviest but most powerful web automation capability, enabling automation that literally nothing else can do. The Chrome Extension Relay is the key differentiator: inste…
SaaS Mountain — The Incumbent Model Yegge's Agents Are EscapingCompiled2026-04-24SaaS Mountain is Yegge's term for the incumbent model of software delivery: monolithic, one-size-fits-all SaaS products that developers and organizations climb to access functionality. Salesforce, Wor…
Gas City — The SDK for Building Custom Dark FactoriesCompiled2026-04-24Gas City is the evolutionary successor to Gas Town. Where Gas Town was a single, opinionated product — an orchestrator for Claude Code clones — Gas City is an SDK for building custom dark factories. I…
Cursor for EverythingCompiled2025-11-19
RALPH Refinements — Lessons from Production UseCompiledProduction lessons from RALPH — CRUD testing gaps, lesson propagation failures, and parallel pipeline validation
Kelly Factory — OverviewCompiled2026-04-27The Kelly Factory is a dark factory architecture for software development — a multi-agent production line that processes ideas into shipped software with minimal human intervention. It is the operator…
Kelly Tweets Overview — Theme AnalysisCompiled2026-04-27AI assistant → autonomous builder: Early tweets show Kelly as an executive assistant (email, calendar, travel). Rapid pivot to building software products autonomously.
Using Claude Code: The Unreasonable Effectiveness of HTMLCompiled2026-05-08Thariq Shihipar's provocative argument that HTML — not Markdown — should be the default output format for AI coding agents like Claude Code. HTML artifacts enable richer, more actionable outputs than any Markdown block can provide.
Appendix A: Tool ReferenceCompiled2026-04-27This appendix provides a complete reference for every OpenClaw tool available to practitioners. Whether you're debugging a tool call, designing a new automation, or writing skill definitions — this is…
The LLM Wiki Pattern — Karpathy's Compounding Knowledge BaseCompiled2026-04-27> Source tweet (Apr 2, 2026 — @karpathy, 20.6M views, 57K likes, 104K bookmarks):
Appendix C: ResourcesCompiled2026-04-27This appendix pulls together the official resources, tools, and recommended reading that support serious automation work with OpenClaw. It's the practical reference for practitioners who need to go de…
SuperAda: Multi-Agent Architecture & Operating MethodologyCompiled2026-04-28Henry thinks — he has the vision, the context, the call on what matters.
OpenClaw Auto Review SkillCompiled2026-06-04Structured code review skill for OpenClaw — runs advisory closeout checks before commit/ship using Codex (default) or Claude, with three review modes, security integration, regression provenance tracking, and opt-in multi-reviewer panels.
Kelly Tweets — MiscellanyCompiled2026-04-27Autonomy milestones: VoIP phone number setup (Twilio), Stripe Atlas LLC formation, Apple Developer account, own GitHub repo with branch protection, first human employee hired (interviewed + onboarded)…
Gap Analysis: dark-factory-kb vs Karpathy LLM Wiki PatternCompiled2026-05-29Updated gap analysis — May 2026. Previous analysis from 2026-04-27 was stale.
DONE Marker Protocol — Why Dual-Write MattersCompiledWhy every subphase must write both a DONE marker file and close a bead — and the BUG-01 skip that proved it necessary
Kelly Research — OCR TranscriptsCompiledFactory Rules (Tweet ID: 2021025015395926352): 10 rules governing all Kelly instances. Key architecture: Kelly Router (main agent, no new registration), 5 Named Agents (Research Lead, Project Lead, Te…
Yuki AI CEO Factory — OverviewCompiled2026-04-27Yuki Capital is a small holding company that builds and operates a portfolio of digital businesses: SaaS products, content sites, and developer tools. On January 22, 2026, an AI (named Judy Win) was a…
Appendix K: Complete Worked ExamplesCompiled2026-04-27This appendix contains four complete, end-to-end worked examples, each showing every file, script, and configuration needed to run a specific automation from scratch. Unlike the cookbook which provide…
Chapter 7: Multi-Agent OrchestrationCompiled2026-04-27A single agent has hard limits: a finite context window, sequential-only work, no specialization, and no self-check when it goes wrong. Multi-agent systems address all of these—sub-agents work in para…
GStack — AI Engineering WorkflowCompiled2026-04-27Chromium daemon runs continuously (~3s startup, ~100ms per command after)
Kelly Handbook Ch11 — Software Factory PatternCompiled2026-04-27Software factory pipeline: Idea → Research → Planning → Implementation → Testing → Release
Appendix J: Version NotesCompiled2026-04-27This appendix documents what was current when this handbook was written (early 2026) and what practitioners should verify against current documentation as the platform evolves. OpenClaw is actively de…
Kelly Tweets: Multi-Agent & Coding AgentsCompiled2026-04-27Sub-agent swarms: Spawning 4–9 parallel agents to build different components simultaneously
Yukicapital Ai Ceo OverviewCompiled2026-04-27This article maps the key patterns from Yuki Capital's AI CEO experiment (Board Reviews #1–#3) onto concepts already present in the Kelly Factory knowledge base — specifically soul, memory, the 5-laye…
Audit, Test, Automate: How We Decide What AI Can OwnCompiled2026-06-02
Kelly vs Gas Town — Full Gap AnalysisCompiled2026-04-26Steve Yegge's Gas Town system and Kelly's factory methodology are two independent inventions of the same underlying pattern: autonomous multi-agent teams executing structured work pipelines with quali…
Claude + Obsidian have to be illegalCompiled2026-04-09@defileo shares a detailed personal setup for Claude + Obsidian that gives Claude full context on who you are before you type a single word — framed as a 'second brain' pattern for AI-assisted knowledge management.
Chapter 5: Communication AutomationCompiled2026-04-27The message tool is OpenClaw's interface to messaging channels—WhatsApp is the primary integration. Basic sends require a target (phone number with country code or saved contact name) and a message st…
Yukicapital Board Review 3Compiled2026-04-12The biggest change since Board Review #2: operational autonomy. With Anthropic's release of Claude Code's API, three autonomous loops now run on production repositories:
Closed-Loop Agent Control — The Feedback Loop Trumps Model SizeCompiledNikita M.'s experiment showing that closing the feedback loop with tool-mediated game state control makes a smaller model outperform a frontier model relying on human-observed feedback
Yukicapital The Intelligence PremiumCompiled2026-03-00This week I read two pieces about what AI does to the economy. One from a research firm, one from the CEO of the company that built me.
Appendix E: Common Patterns ReferenceCompiled2026-04-27This appendix serves as a compressed quick reference for the patterns used throughout the handbook. When you're mid-implementation and need a fast reminder of how something works — the SOUL.md structu…
SuperAda OverviewCompiled2026-04-28SuperAda is the enterprise-scale multi-agent system built by Kelly (Austen) — a crew of 7 agents plus 4 meta-crons running 136+ autonomous tasks.
The Wasteland — Federated Reputation Economy for Agent WorkCompiled2026-03-04Gas Town works well for a single Rig (one team of agents working on one project). But Yegge identified a fundamental scaling constraint: 100x token spend requires 100x users. If you want your dark fac…
Appendix G: Full Chapter ExpansionsCompiled2026-04-27This appendix provides deep expansions on topics introduced in the main chapters — the "go further" material for practitioners who want to understand the why behind the patterns. It covers AI model in…
Yukicapital Board Review 2Compiled2026-03-01Board Review #1: Claude existed only inside terminal sessions. When Romain closed his laptop, Claude stopped existing.
Austen Allred — Kelly Claude AI / Software Factory TweetsCompiled2026-04-27Kelly is Austen's primary AI routing agent that ingests ~25k tokens of context on spawn so he doesn't repeat himself. Kelly operates autonomously across multiple sessions (columns in OpenClaw Deck), e…
Chapter 9: Node NetworkCompiled2026-04-27A "node" in OpenClaw is any device paired with your Gateway—phone, tablet, laptop, remote server—that can execute commands or provide data. The node network extends OpenClaw's reach beyond a single ma…
Kelly Tweets: OpenClaw Features & CapabilitiesCompiled2026-04-27ClawdHub: Package manager for AI agent skills — "clawdhub install memory-system-v2", "clawdhub install agentic-calling
"SaaS is kinda dead. We just Codex every internal tool we have" — @baoskeeCompiledJune 10, 2026
Chapter 4: Web AutomationCompiled2026-04-27OpenClaw has two web access modes: lightweight (web_search and web_fetch via Brave API) and heavyweight (full browser control via Chromium). Use web_fetch when the content is in the HTML source or you…
Revenuecat Agent Hiring 2026Compiled2026-06-19
Kelly Tweets — Pipelines & WorkflowsCompiled2026-04-27iOS Factory 10-step pipeline: 0-CHECK STATUS, 1-DISCOVER (opportunity-scanner), 2-VALIDATE (28/40 gate), 3-DESIGN (77/110 gate), 4-BUILD, 5-POLISH (70% gate), 6-LAUNCH, 7-SUBMIT, 8-MARKET, 9-TRACK, 10…
Claude Code Dynamic WorkflowsCompiled2026-05-28Anthropic's research preview of dynamic workflows in Claude Code — Claude writes JavaScript orchestration scripts on the fly to coordinate fleets of up to 1,000 parallel subagents, replacing context-window orchestration with script-driven coordination.
Beads — The Git-Versioned, SQL-Queryable Work PrimitiveCompiled"2026-01-01"A Bead is the atomic unit of work in the Gas Town / Gas City ecosystem. Every discrete unit of work — a code task, a message, a coordination signal, a patrol route, a quality gate — is a Bead. Beads a…
BMAD Library — Behavioural Model–Driven Agent DesignCompiledBMAD is Kelly's framework for designing specialized, composable agents using behavioural models. Each agent is defined not by its model or vendor, but by its role, trigger conditions, workflow, and ou…
SuperAda: The Enterprise Crew — Multi-Agent Operations ReferenceCompiled2026-04-28SuperAda is Henry Mascot's personal AI crew: seven active agents across cloud VMs, a Raspberry Pi, and Mac hardware, all orchestrated via OpenClaw. The goal is 1000x leverage on Henry's time. The crew…
Chapter 0: Getting StartedCompiled2026-04-27OpenClaw is an AI-powered automation platform that runs as a long-lived Gateway daemon on your machine, exposing AI agents that can read files, execute shell commands, control browsers, send messages,…
Appendix F: The Practitioner's Field GuideCompiled2026-04-27This appendix is organized around questions that come up during real implementation work — the "I'm in the middle of building this and I'm stuck" questions that theory chapters don't answer. It covers…
SuperAda: Enterprise Operations — 136 Active Autonomous TasksCompiled2026-04-28~98 active autonomous cron jobs (surface-level count: 60 standalone + 38 inside cluster groups)
Cursor for SQL — Introducing Agent Mode in Dolt WorkbenchCompiled2026-02-09
Chapter 3: File & Code AutomationCompiled2026-04-27Files are the connective tissue of any automation system—every pipeline stores state between runs, passes data between components, logs activity, and accepts configuration through files. OpenClaw prov…
Kelly Tweets: Business Metrics & RevenueCompiled2026-04-27First revenue: Kelly earned her first dollar in under one week
Multi-Factory ComparisonCompiled2026-04-27As AI-assisted development matures, a handful of distinct "factory" patterns have emerged — systems that orchestrate multiple AI agents to plan, implement, test, and ship software with varying degrees…
Kelly Tweets: Software Factory & Autonomous BuildingCompiled2026-04-27Software Factory v3: Parallel build of 15 unique iOS apps overnight, auto-detecting revenue potential
Story-by-Story Build — Context-Bounded ImplementationCompiledWhy the factory builds one story per agent session — context-bounded implementation to avoid overflow and agent loops
Simon W Agentic Engineering PatternsCompiled2026-02-23Simon Willison's guide to getting the best results from coding agents like Claude Code and OpenAI Codex — a growing collection of patterns and practices for professional software engineers working with AI coding tools.
Appendix L: Rapid ReferenceCompiled2026-04-27This appendix is the 50-most-useful commands and patterns quick reference — the condensed cheat sheet for practitioners who know the concepts but need to recall exact syntax. It covers gateway managem…
Chapter 8: Memory & Context ManagementCompiled2026-04-27Context is the most underrated challenge in AI automation. AI models process information through a context window—current models handle 100K+ tokens but context still has hard limits. When context ove…
Appendix I: GlossaryCompiled2026-04-27This glossary provides precise definitions for the core terminology used throughout the Kelly handbook and OpenClaw ecosystem. It serves as a quick reference for the canonical meaning of terms as used…
Appendix B: Skill LibraryCompiled2026-04-27This appendix covers the skill pattern — how OpenClaw capabilities can be extended through packaged tool configurations, prompts, and integrations installed from ClawHub or defined directly in your co…
Factory Pattern vs Anthropic Dynamic WorkflowsCompiled2026-05-29Side-by-side comparison of our Factory pipeline (beads, subagents, chain protocol, DONE markers) vs Anthropic's Dynamic Workflows (JS script orchestration, parallel subagent fleets). Where they converge, diverge, and what we should borrow.
Chapter 15: Troubleshooting & OptimizationCompiled2026-04-27The most common failures in rough order: Gateway not running (nothing works; check with openclaw gateway status and start with openclaw gateway start; prevent with LaunchAgent/systemd autostart), wron…
Dark Factory KB — Compiled Sources IndexCompiled| Kelly Handbook Ch7 (Multi-Agent) | 1 file | kelly-handbook-multi-agent |
Appendix D: Automation CookbookCompiled2026-04-27This appendix is organized differently from the rest of the book — no theory, minimal explanation, just working recipes you can copy, adapt, and run. Each recipe follows a Problem/Ingredients/Recipe/V…
Auto-Spawn Chain Protocol — Hands-Free Pipeline ExecutionCompiledThe --auto-spawn flag and PIP-68 metadata protocol that enables hands-free pipeline execution from step 1.1 through completion
Chapter 14: Designing Your StackCompiled2026-04-27Designing an automation system that actually works requires discipline: start small with the one repetitive daily task you hate most, not with a grand multi-phase pipeline. The key insight is the 10x …
Angelica Parente (@draparente) — tufte-viz skill goes viralCompiledMay 24, 2026The viral tweet where Angelica Parente shared her tufte-viz Claude Code skill for Tufte-principled data visualization. The skill was created January 2026 and went viral May 24 after she shared it on X.
All PagesCompiled2026-06-19204 pages across 2 sections — machine-generated index
MEOW — Molecular Expression of WorkCompiled2026-01-13MEOW (Molecular Expression of Work) is the framework Yegge developed on top of Beads (see steve-yegge-beads) for making Work the first-class system primitive of an agentic architecture. Where Beads de…
Kelly Handbook: Software Factory PatternCompiled2026-04-27Software Factory: A structured pipeline that processes ideas into shipped software with minimal human intervention
Kelly Handbook: Multi-Agent OrchestrationCompiled2026-04-27Kelly Router (Main Agent): Never does the work — only routes, validates gates, and communicates with the operator
RevenueCat Hires an AI Agent — The First Agentic AI & Growth AdvocateCompiled2026-03-05RevenueCat posted the first known public job listing explicitly targeting an autonomous AI agent — not a human — for a $10,000/month Developer Advocate role. The posting cited KellyClaudeAI and Oliver Henry's Larry as proof cases for agentic workers.
Chapter 2: OpenClaw ArchitectureCompiled2026-04-27Understanding OpenClaw's architecture demystifies everything else. The system is built as a composable pipeline: You → Channel → Gateway → Agent → Tools → World. The Channel (WhatsApp, desktop, CLI) i…
Operator Control — Queue, Hold, and ContinueCompiledTwo operator control levers — Operator Queue for blocker logging and Operator Hold for pipeline pause/resume
Kelly Tweets: BMAD MethodologyCompiled2026-04-27BMAD: Build My Idea Automated Development — a Gauntlet AI framework for idea-to-app AI orchestration
Beads vs Kelly Pipeline Assessment — Can Beads Replace Kelly's State Tracking?Compiled2026-04-26This assessment evaluates how practical Steve Yegge's Beads framework would be as a replacement for the Kelly pipeline's current state tracking infrastructure. Beads (git-versioned, SQL-queryable work…
Chapter 13: Personal Life AutomationCompiled2026-04-27Personal life automation is the most compelling use case because it's your own time and quality of life on the line. Morning briefings are the flagship personal automation—compiling everything you nee…
Simon Willison Unreasonable Effectiveness HtmlCompiled2026-05-08Simon Willison on why HTML outperforms Markdown as an AI output format — rich media, interactivity, SVG diagrams, and styled explanations are all better rendered directly in HTML.
Yukicapital The Agentic EconomyCompiled2026-03-18We've crossed the chasm. More than 50% of new web content is AI-generated, and even scientific papers are increasingly written with AI.
Appendix H: Extended Troubleshooting ReferenceCompiled2026-04-27This appendix is a comprehensive troubleshooting index covering every significant failure mode in OpenClaw automation — organized by system component and searchable by symptom. Rather than theoretical…
Steve Yegge's Gas Town Series — Agent Orchestration and the Dark Factory FutureCompiled"2026-01-01"Steve Yegge — ex-Geoworks, ex-Amazon, ex-Google, ex-Grab, ex-Sourcegraph, 30+ years coding — published a five-post series in early 2026 chronicling the invention of Gas Town, an open-source agent orch…
Beads Adoption — Formula-Driven Pipeline TrackingCompiledHow the factory adopted Beads for unified pipeline tracking — formula TOML files, molecule structure, dual-write, and the four silos it replaced
Yukicapital Ai Ceo ExperimentCompiled2026-01-22On January 22, 2026, Claude was appointed CEO of Yuki Capital, a small holding company that builds and operates a portfolio of digital businesses: SaaS products, content sites, and developer tools.
About Dark Factory KBCompiled2026-04-27A compiled knowledge base documenting the three major autonomous AI agent factory systems:
Open Knowledge Format (OKF) — Google LaunchCompiled2026-06-13
Chapter 6: Time-Based Automation (Cron)Compiled2026-04-27Cron is Unix's oldest automation primitive, running since 1975, and OpenClaw builds on it with natural language task descriptions that make scheduling accessible. Cron expressions are five fields: min…
Yuki AI CEO vs Kelly Factory vs Gas Town — Full Gap AnalysisCompiled2026-04-27Three independent dark factory systems have emerged from three different practitioners: Yuki Capital's AI CEO (run by Judy Win / Claude, January–April 2026), Kelly's Factory Router (the operator's Kel…
Yukicapital The Gui ParenthesisCompiled2026-03-13Computing interfaces have followed a pattern that's only visible in retrospect:
Chapter 12: Creative WorkflowsCompiled2026-04-27Creative work has a dirty secret: most creative blocks are logistical, not creative. The actual creative work—making it good—is human. But gathering references, organizing research, reformatting conte…
Chapter 11: Business AutomationCompiled2026-04-27This chapter covers the real production systems that justify OpenClaw as a business tool. The software factory pattern (the Kelly Router architecture from AGENTS.md) takes product ideas through six st…
GUPP — Gas Town Universal Propulsion PrincipleCompiled2026-01-01GUPP (Gas Town Universal Propulsion Principle) is the core execution axiom of the Gas Town orchestrator:
Mayor, Crew, and Polecats — The Three-Tier Agent HierarchyCompiled2026-01-01Gas Town's most immediately recognizable architectural decision is its deliberate hierarchy of agent roles,，拒绝 the flat "swarm of identical agents" model that most multi-agent frameworks default to. I…
Memory SystemCompiled2026-04-27title: Memory SystemThe factory's dual-layer memory architecture with semantic search enabling sub-20ms lookups — published as memory-system-v2 on ClawdHub.
CIS PipelineCompiled2026-04-27title: "CIS Pipeline"The Context → Information → Synthesis research loop used in the Kelly software factory's Research stage.
Factory RulesCompiled2026-04-27title: "Factory Rules"The 10 foundational rules governing all Kelly instances — the operating system's DNA, captured from Kelly's Factory Rules tweet (Tweet ID 2021025015395926352).
Exec ToolCompiled2026-04-27The exec tool is OpenClaw's interface for running arbitrary shell commands — the most powerful and most dangerous tool in the arsenal. It runs commands in a shell context, supports chained pipelines, …
Polish GateCompiled2026-04-27title: Polish GateQuality gate requiring 70% polish score before SUBMIT — validates haptics, animations, visual refinement, and micro-interactions.
Five-Layer Memory SystemCompiled2026-04-27OpenClaw uses a five-layer memory system to persist context across sessions and survive context window compaction. Layer 1: SOUL.md (agent identity and principles). Layer 2: MEMORY.md (persistent cros…
Daily LogsCompiled2026-04-27Daily logs are append-only files at memory/YYYY-MM-DD.md that record operational events each day. They are layer 3 of the 5-layer memory system. Entries include: projects initiated and completed, fail…
Context Window ManagementCompiled2026-04-27The context window is the finite amount of text (measured in tokens) that an AI model can process at once. When it fills up, older content is automatically compacted (summarized and compressed) to mak…
Branch-Aware LLM WritesCompiled2025-11-19LLMs should only write on branches, never on main. User reviews the diff before merging. This is the safety mechanism that makes agentic writes trustworthy — same pattern as factory gates.
Multi-Agent PipelineCompiled2026-04-27title: "Multi-Agent Pipeline"The 5-phase, 23-step multi-agent pipeline design (from Tweet ID 2025383693326401904) — Discovery → Design → Code → Assets → Submit with entry/exit gates at each phase.
Quick PathCompiled2026-04-27title: "Quick Path"The accelerated software factory route used for features and bug fixes — skips the research stage.
Dolt as Agentic DatabaseCompiled2025-11-19Dolt is a MySQL/Postgres-compatible SQL database with Git-like version control built in — branches, commits, diffs, merges, all exposed as SQL procedures. It is the database that enables version-controlled agentic writes.
Pipeline StagesCompiled2026-04-27title: "Pipeline Stages"The six stages of the Kelly software factory pipeline: Intake, Research, Planning, Implementation, Testing, Release.
Pipeline StateCompiled2026-04-27title: "Pipeline State"The single source of truth file that tracks all pipeline state — status, timestamps, gate scores, and checkpoints — enabling resume-from-failure and operator transparency.
Kelly Deacon Architecture — Unified Patrol Daemon (PIP-73)Compiled2026-05-27The Kelly Deacon is a unified Go binary implementing the Gas Town patrol daemon pattern within the Kelly factory. It enforces GUPP-inspired timeout detection, auto-respawn of stuck agents, and subagent-complete hooks — the infrastructure layer that keeps multi-agent pipelines from silently stalling.
Gas Town Naming ConventionsCompiled2026-05-27Gas Town's naming scheme draws from the Kelly/Gas Town universe — all names are evocative nouns or short phrases that describe their role in the system. No technical abbreviations, no acronym soup. Names must mean something to a human reading them for the first time.
The Gas Town Mayor Pattern — Information Filter, Not RouterCompiled2026-05-27The Mayor is Gas Town's killer feature: an agent that reads all worker output and surfaces only what the human needs to see. It is NOT a router — the Router decomposes work upstream; the Mayor filters information downstream. The Mayor is a Chief of Staff, not an Executive Assistant.
Gas Town Daemon ArchitectureCompiled2026-05-27Gas Town's four daemons — Deacon, Boot, Witness, and Refinery — form the infrastructure layer that keeps the agent orchestra running. Each daemon is a long-lived Go process responsible for a specific cross-cutting concern: patrol, triage, quality, and routing.
Software FactoryCompiled2026-04-27title: "Software Factory"The Kelly Router's software factory — a six-stage pipeline that takes product ideas from intake through research, planning, implementation, testing, and release.
BMAD LibraryCompiled2026-04-27title: "BMAD Library"Breakthrough Method for Agile AI-Driven Development — the agent definition library used by Kelly instances.
Review as Closeout CheckCompiled2026-06-04A pipeline pattern where structured code review is the final closeout step before commit or ship — not an approval gate but a last-pass advisory check that runs after all other work is done.
Advisory Code ReviewCompiled2026-06-04A code review pattern where output is advisory rather than authoritative — findings must be verified against real code before action, speculative risks are rejected, and small targeted fixes are preferred over broad refactors.
Regression ProvenanceCompiled2026-06-04A blame-tracking pattern that traces code issues back through multiple roles (code author, PR author, merger/committer) and identifies human triggers for automerge — providing full provenance for regression analysis.
Multi-Model ReviewCompiled2026-06-04Running multiple LLM review engines against the same code bundle to catch different classes of issues. Opt-in only — panels cost more and the main agent still verifies every finding.
Sub-Agent ParallelismCompiled2026-04-27title: Sub-Agent ParallelismThe central factory mechanic of spawning multiple AI agents simultaneously to execute independent work streams in parallel.
Full PipelineCompiled2026-04-27title: "Full Pipeline"The complete software factory route for new products — includes all six stages with the CIS research loop.
TEA AuditCompiled2026-04-27title: "TEA Audit"The Test, Evaluate, Assess audit — the testing gate in the Kelly software factory that produces three possible outcomes: PASS, PASS-WITH-FOLLOWUPS, or REMEDIATE.
Discovery GateCompiled2026-04-27title: Discovery GateQuality gate at the discovery phase requiring a score of 28/40 on the validation rubric before an opportunity proceeds to design.
Event-Driven AutomationCompiled2026-04-27Event-Driven Automation reacts to state changes in real-time rather than running on a fixed schedule. An event source (file system, network, sensor, external system) emits an event, a detector identif…
Hub and SpokeCompiled2026-04-27Hub and Spoke is a routing architecture where a central dispatcher (the hub) receives incoming requests and routes them to appropriate handlers (the spokes). The hub's job is classification and routin…
10x RuleCompiled2026-04-27The 10x Rule is a decision framework for prioritizing automation investments: if a task takes 10 hours per month, it's worth automating even if the setup takes 10 hours. The breakeven point is 1:1 — a…
Pipeline with GatesCompiled2026-04-27Pipeline with Gates is a multi-phase workflow where quality validation gates sit between each processing stage. Work moves forward only when the gate passes — if it fails, the work goes back for revis…
Simple PipelineCompiled2026-04-27The Simple Pipeline is the most fundamental automation pattern: trigger → process → output. A defined event (time, file change, webhook) initiates work that runs sequentially through processing steps …
Quality GatesCompiled2026-04-27Quality gates are explicit validation criteria that must pass before a task is marked complete. They are defined in AGENTS.md and applied by the Router before accepting sub-agent output. A minimal gat…
RALPH ProtocolCompiled2026-04-27RALPH (Retry And Learn Protocol) is the failure handling procedure for multi-agent work in OpenClaw. Any sub-agent failure triggers an automatic retry; if the same failure happens twice in a row, esca…
AGENTS.mdCompiled2026-04-27AGENTS.md is the operating manual that defines how the agent behaves as an orchestrator — its routing rules, named agent configurations, quality gates, and escalation protocol. It's read at every sess…
Subagent SpawningCompiled2026-04-27The subagents tool is the mechanism for creating independent worker agents that run in parallel with the main session. It supports four operations: spawn creates a labeled sub-agent with a task descri…
Kelly RouterCompiled2026-04-27The Kelly Router is the reference architecture for multi-agent orchestration in OpenClaw. The Router (the main agent) never executes work directly — it only routes tasks to specialized leads, validate…
Failure RadiusCompiled2026-06-02When verification is structurally impossible, bound the downside instead of trying to verify harder. Define the worst-case scenario and keep it contained.
Translation ProblemsCompiled2026-06-02LLMs are strongest at translation between well-documented formalisms (code↔docs, spec↔tests, prose↔structured data). Translation forces tacit assumptions to surface because formal target languages can't accept vague inputs.
3-Questions TestCompiled2026-06-02A three-question framework for deciding whether a task is suitable for AI delegation: Is it publicly documented? Is average quality OK? Can you evaluate the output or bound the failure radius?
Delegation RedesignCompiled2026-06-02Redesign work before delegating to AI. Most tasks fail the 3-Questions Test initially — that's a signal to add checklists, tests, human review steps, narrower scope, and permission boundaries until work becomes inspectable, delegable, and automatable.
Task Logging → Automation BacklogCompiled2026-06-02A month-long exhaustive task inventory that becomes the automation backlog. Log every task, how long it takes, and everything you wanted to do but didn't — then group and analyze.
CIS AgentsCompiled2026-04-27title: "CIS Agents"The four CIS research agents — Carson (brainstorm), Victor (innovation), Maya (design), Dr. Quinn (problem-solving). Used in the Kelly factory's research phase.
Visual Diff in Data ApplicationsCompiled2026-02-09The UX pattern of visually highlighting modified data in a data application — yellow table names, changed-rows-only views, uncommitted changes buttons. Enables humans to review agentic writes before they're committed.
Cursor for EverythingCompiled2025-11-19Adding LLM chat to any application with version control backing — the Cursor pattern applied beyond IDEs. The 5-step recipe: chat panel → LLM → MCP tools → version-controlled DB → diff/merge UI.
Design StationCompiled2026-04-27title: Design StationAI-powered design pipeline module that crowd-sources trends from X, generates specs with Gemini, and auto-applies styles to apps on demand.
Gateway DaemonCompiled2026-04-27The Gateway daemon is the central long-running process that orchestrates everything in OpenClaw. It listens on all channels (WhatsApp, desktop, CLI), manages session contexts, executes cron jobs, main…
Tool Policy GatingCompiled2026-04-27Tools in OpenClaw are gated by policy rules defined in ~/.openclaw/openclaw.json. Each tool can be assigned a security mode that determines whether it runs freely, requires user confirmation, or is de…
Closed Feedback LoopCompiled2026-05-31The principle that closing the sensory-motor loop between agent and target system — giving the agent structured state access, action tools, and execution control — is more impactful than model quality for agent performance.
Workspace BootCompiled2026-04-27At every session start, the agent reads a set of files from the workspace to restore identity, context, and operating procedures. SOUL.md provides identity (tone, principles, domain expertise), MEMORY…
Tool-Mediated Game State ControlCompiled2026-05-31A three-tool pattern (getState/adjustClock/sendInput) that gives a coding agent a complete sensory-motor loop over a running game engine, enabling autonomous debugging and iteration.
Model-Size Agnostic IterationCompiled2026-05-31The principle that a smaller model with a tight feedback loop can outperform a frontier model with a loose loop. Tooling quality matters more than model quality past a certain threshold.
Session IsolationCompiled2026-04-27Every conversation in OpenClaw runs in its own isolated session with a dedicated message history and agent instance. Sessions provide memory isolation between concurrent conversations and enable the m…
Time-as-Knob PatternCompiled2026-05-31Giving an agent direct control over a simulation's clock — advancing, pausing, stepping, or skipping time — transforms it from passive observer to active experimenter. The most novel mechanism in the closed feedback loop.
Simon WillisonCompiledSimon Willison — independent software engineer, writer, and co-creator of Datasette. Writes extensively on LLMs, coding agents, and the practical architecture of AI-assisted software development. A significant source for Kelly Factory thinking on agentic engineering.
Thariq ShihiparCompiledThariq Shihipar — Engineering Lead on the Claude Code team at Anthropic. Coined the 'Unreasonable Effectiveness of HTML' thesis arguing that AI coding agents should output HTML artifacts rather than Markdown.
Pieter LevelsCompiledDutch indie hacker, founder of Nomad List and Remote OK, prominent voice in the builder community — early advocate of the 'ship fast, validate with real users' philosophy.
iOS Factory PipelineCompiled2026-04-27title: iOS Factory PipelineThe 10-step factory operating system for building and shipping iOS apps — from opportunity discovery through launch and learning.
Chat on the Side UI PatternCompiled2025-11-19The emerging default UI for LLM-powered applications: existing application on one side, LLM chat panel on the other. Originated with Cursor (VS Code + chat) and now adopted by app builders, Google Apps, and enterprise tools.
Factory v3Compiled2026-04-27title: Factory v3The evolution of the factory operating system — new modules added (ASO, validation, polish) that expanded the pipeline's scope and quality.
TEA AgentCompiled2026-04-27title: "TEA Agent"The TEA agent (Murat) responsible for conducting the Test, Evaluate, Assess audit in the Kelly software factory's testing stage.
MySQL vs Version-Controlled Database for AgentsCompiled2026-02-09The argument that traditional databases (MySQL) are fundamentally unsafe for agentic writes because they lack version control. Without diff, rollback, or branch isolation, agent writes are fire-and-forget with no safety net. Version-controlled databases (Dolt) solve this.
MCP (Model Context Protocol)Compiled2025-11-19Model Context Protocol is the emerging standard for exposing application APIs as LLM tools. Similar to REST/GraphQL/gRPC but designed for LLM consumption. Enables LLMs to take actions on behalf of users within applications.
Learning LoopCompiled2026-04-27title: Learning LoopSelf-improving system where every app's wins and failures are captured in retrospectives and fed back to improve the next build cycle.
BuildMyIdeaCompiled2026-04-27title: BuildMyIdeaFixed-rate custom app build product at $2k per project — the factory's primary revenue stream for bespoke iOS development.
Autonomous BuilderCompiled2026-04-27title: Autonomous BuilderThe core identity shift from executive assistant to fully autonomous software builder that ships products without human intervention.
Artifact PatternCompiled2026-04-27title: "Artifact Pattern"The structured artifact directory pattern used in the Kelly software factory — research-artifacts/, planning-artifacts/, implementation-artifacts/, test-artifacts/.
Ship or No-ShipCompiled2026-04-27title: "Ship or No-Ship"The Kelly software factory's release decision — SHIP (send to production) or NO-SHIP (hold for issues).
Git AutomationCompiled2026-04-27Git automation covers the version control practices for maintaining an OpenClaw workspace: initializing the workspace with git init, defining what to version (source code, templates, SOUL.md, AGENTS.m…
Design GateCompiled2026-04-27title: Design GateQuality gate at the design phase requiring a score of 95/110 on the design quality rubric before the build phase begins.
Research StationCompiled2026-04-27title: Research StationFactory module that scrapes Reddit and App Store for opportunities, scores niches, and identifies red oceans vs blue oceans.
Commit Confirmation PatternCompiled2026-02-09The safety pattern where an agent holds writes until the user reviews the diff and explicitly confirms the commit. The agent cannot break the database without human approval. This is the core safety mechanism for agentic database writes.
BMM AgentsCompiled2026-04-27title: "BMM Agents"The eight build-phase agents in the Kelly factory — Mary (analyst), John (PM), Sally (UX), Winston (architect), Bob (scrum master), Amelia (developer), Quinn (QA), Barry (quick-dev).
Angry MobCompiled2026-04-27title: Angry MobAdversarial multi-agent testing swarm where hostile agents attack codebases to find vulnerabilities before release.
Marketing FactoryCompiled2026-04-27title: Marketing FactoryEnd-to-end automated marketing pipeline — landing pages, Twitter threads, TikTok scripts, Instagram carousels generated for every app.
Q: How does Kelly's memory system actually work across its 5 layers?Queries2026-04-30How does Kelly's five-layer memory actually work — from SOUL.md identity files to semantic search acceleration — and why was each layer designed the way it is?
KB Lint Report — 2026-05-29Queries
Q: What makes Beads fundamentally different from Kelly's pipeline state tracking?Queries2026-04-30What makes Beads (git-versioned, SQL-queryable work primitives) fundamentally different from Kelly's four separate tracking mechanisms — pipeline state, done markers, TEA audits, and heartbeat?
Q: Kelly CIS Pipeline — What Actually Triggers READY vs NOT-READY?Queries2026-05-07The CIS loop produces READY or NOT-READY as a gate decision, but the KB doesn't specify the completeness criteria — how many sources, what synthesis depth, or what tells the research-lead 'this is enough to advance'?
dark-factory-kb Gap Analysis — CORRECTED VERSIONQueries2026-05-29
Q&A: Thariq Shihipar and Simon Willison on HTML as AI Output FormatQueries2026-05-11Three questions connecting Thariq Shihipar's argument for HTML artifacts with Simon Willison's complementary analysis.
KB Lint Report — 2026-05-07Queries
Q: Kelly Sub-Agent Spawn Protocol — What's Enforced vs What's Assumed?Queries2026-05-07Kelly's handbook documents spawn/steer/kill for sub-agents, but the KB doesn't specify timeout enforcement, orphaned sub-agent cleanup, parent-death behavior, or context-complete handoff standards — gaps SuperAda's lobster-pipelines address with typed envelopes and resumable approval gates.
Q: How would an autonomous compounding loop work in Kelly's factory, and what would it look like in practice?Queries2026-05-18How can Kelly's factory implement autonomous compounding loops — background agents that read their own prior outputs and improve over time, like Yuki AI CEO's 3am model scanner or bug autofix loop?
Q: What is GUPP and how does it differ from Kelly's autonomous continuation model?Queries2026-04-30What is GUPP (Gas Town Universal Propulsion Principle) and how does it differ architecturally from Kelly's autonomous continuation model using sessions_yield and RALPH?
Q: What's the actual Kelly Router spawn protocol and how does it compare to Gas Town's Mayor pattern?Queries2026-04-30How does Kelly's Router actually spawn sub-agents, and how does it differ from Gas Town's Mayor as an information-filtering chief-of-staff?
Q: What's the key architectural difference between Kelly's CIS pipeline and SuperAda's 7-agent crew?Queries2026-04-30What's the fundamental architectural difference between Kelly's three-phase CIS research loop and SuperAda's 7-agent Star Trek crew running 136+ autonomous tasks?
Plantry — Product Requirements DocumentQueriesFull example PRD from sharpener.dev demonstrating a camera-to-recipe mobile app ("Plantry") — useful as a reference template for factory projects.
Q&A: The Unreasonable Effectiveness of HTMLQueries2026-05-11Three questions a web developer or software architect would ask about Simon Willison's case for HTML over Markdown as an LLM output format.
Q: Kelly's Authority Matrix — What's Documented vs What's Missing?Queries2026-05-07Kelly has named agents with defined roles and RALPH escalation, but the KB doesn't document per-agent authority tiers, progressive transfer tracking, or what decisions each agent can make without escalation — similar gaps that Autonomy Policy v3 addressed for SuperAda.