Agentic Knowledge Wiki Index

A private, markdown-native research and synthesis space for agentic development, local models, orchestration patterns, and tool/runtime evolution.

Start here

• Now — shortest path to what matters right now
• Best Of — strongest pages in the repo
• Decision Dashboard — current decisions and near-decisions
• Recently Added — newest important additions
• QUA-260 Outstanding Queue — what is extracted vs what still needs real assessment work
• Overview — what this wiki is and how to use it
• Log — chronological changes and ingest history
• QUA-260 First Seed Wave — first major synthesis pass
• QUA-260 Second Seed Wave — second synthesis pass

Fast paths

If you want core operating doctrine

• Harness-First Agent Development
• Planner Executor Separation
• Artifact-First Handoffs
• Context Manifests and Governed Retrieval
• Two-Pass Context Management

If you want the big picture

• Harness-First Agent Development
• Agent Ecosystems vs Single Agents
• Memory vs RAG vs Wiki
• AI Product R&D vs Roadmap Execution
• Solo Operator Stack

If you want practical operating patterns

• Delegation Loop
• Planner Executor Separation
• Artifact-First Handoffs
• Guardrails as Defaults
• Two-Pass Context Management
• Context Manifests and Governed Retrieval

If you want source material behind the current thesis

• OpenAI Harness Engineering
• OpenClaw Operating Model Gap Assessment
• AI Dev Code Builder Agent Topology
• GPT-5.4 1M Context Window Operating Policy
• Everything is Context — Critical Assessment
• Beyond the Prompt — Richard White

If you want strategic/background signals

• AI Product R&D vs Roadmap Execution
• China AI Deployment Gap + Solo Operator Positioning
• Solo Operator Stack

Library map

Concepts

• Compiled Knowledge Systems — why a wiki can beat query-time-only retrieval
• Memory vs RAG vs Wiki — role boundaries between continuity, retrieval, and synthesis
• Harness-First Agent Development — why harness quality matters more than model-chasing alone
• Shared Memory vs Compiled Wiki — where shared operational memory ends and markdown synthesis begins
• Agent Ecosystems vs Single Agents — why sticky systems are ecosystems, not isolated chat wrappers
• AI Product R&D vs Roadmap Execution — why AI-heavy work often behaves more like exploration than classic roadmap delivery
• Solo Operator Stack — reasoning, action, governance, and output as a leverage stack

Patterns

• Incremental Ingest — one-source-at-a-time compilation workflow
• Linting and Drift Control — how to keep the wiki from drifting into polished nonsense
• Planner Executor Separation — keep planning and execution from bleeding into each other
• Guardrails as Defaults — encode repeated lessons into defaults and mechanical checks
• Artifact-First Handoffs — use durable artifacts instead of mushy conversational handoffs
• Delegation Loop — onboard, delegate, review, capture, and orchestrate repeated agent work
• Two-Pass Context Management — go wide for orientation, then narrow hard for execution
• Context Manifests and Governed Retrieval — make context selection explicit and policy-constrained

Runtimes

• OpenClaw Runtime Fit — fit assessment for OpenClaw, memory, corp KB, and corp-strategy-work

Sources

Foundational

• Karpathy LLM Wiki Gist — blueprint for the compiled-knowledge approach
• QUA-260 AI Dev Practices Tracker — Linear umbrella issue acting as the seed queue

Harness and operating model

• OpenAI Harness Engineering — external harness-first engineering signal
• Agentic Harness Statement v1 — internal three-layer harness framing
• Bespoke CLI for Agents Pattern — why repeated messy sources often deserve a CLI + skill instead of raw-context reuse
• OpenClaw Operating Model Gap Assessment — operating-model hardening recommendations
• AI-first Operating Model from Voice Memo — founder/operator strategy note on AI-first execution

Memory, topology, and ecosystem design

• OpenBrain Shared-Memory Pattern — shared database + MCP memory architecture pattern
• AI Dev Code Builder Agent Topology — dedicated build-lane topology for coding work
• The Coolest Agents I've Built So Far — field report on useful agent ecosystems
• Factory Missions Breakout — planning-first orchestration, milestone validation, and targeted parallelism in a production coding-agent system
• GPT-5.4 1M Context Window Operating Policy — tiered context policy for GPT-5.4 and similar workflows
• Knowledge Extraction + AI / Genetic Workflow Integration — broad framing on knowledge extraction into action
• Everything is Context — Critical Assessment — narrow and govern file-system-style context architecture
• Beyond the Prompt — Richard White — AI product development as exploratory R&D under uncertainty
• The Art of AI Delegation — delegation as the human operating skill behind AI leverage
• China AI Deployment Gap + Solo Operator Positioning — strategic background signal on deployment and positioning

Queries

• QUA-260 First Seed Wave — first synthesis pass from the AI Dev tracker
• QUA-260 Second Seed Wave — second synthesis pass focused on topology, handoffs, and context policy
• QUA-260 Outstanding Queue — what is now extracted versus what still needs real assessment work

Decisions

_No decision records yet._

Projects

_No project pages yet._

Tools

• Agent-Facing CLIs — thin command surfaces that make repeated agent workflows more composable, deterministic, and approval-aware

Models

_No model pages yet._

Frameworks

• Factory Missions — planning-first orchestration for long-running autonomous coding with milestone validation and targeted parallelism

Reading suggestions

10-minute orientation

1. Overview
2. Harness-First Agent Development
3. QUA-260 First Seed Wave

Operator playbook path

1. OpenClaw Operating Model Gap Assessment
2. Planner Executor Separation
3. Guardrails as Defaults
4. Artifact-First Handoffs

Architecture path

1. OpenAI Harness Engineering
2. Agentic Harness Statement v1
3. Agent Ecosystems vs Single Agents
4. Shared Memory vs Compiled Wiki