The Knowledge Base

Why a Knowledge Base Matters

My debugger agent spent 40 minutes solving a Next.js hydration error. Two weeks later, the same error appeared and the agent spent another 35 minutes rediscovering the same fix. That 35-minute repeat was the moment I built the knowledge base. Now every non-obvious solution gets captured once and reused forever.

For your business, this means domain expertise persists. Solutions to problems are captured once and reused forever. The system gets smarter over time instead of starting from zero each session.

Knowledge Index — The Master Map

The knowledge index is a single file that catalogs all knowledge files in your system. Agents check this index before doing research — if the answer already exists, they use it instead of searching again.

# Knowledge Index

## System Configuration
| File | Description | Updated |
|------|-------------|---------|
| ~/.claude/CLAUDE.md | Core AI identity | 2026-03-11 |
| ~/.claude/rules/*.md | Behavioral rules | 2026-03-11 |
| ~/.claude/protocols/*.md | Workflow protocols | 2026-03-11 |

## Domain Knowledge
| File | Description | Updated |
|------|-------------|---------|
| ~/life/docs/[topic].md | [Description] | [Date] |

## Skills (Reusable Patterns)
| File | Pattern | When to Use |
|------|---------|-------------|
| ~/.claude/skills/[skill].md | [Pattern name] | [Trigger condition] |

Skill Files — Reusable Patterns

Skills are small, focused files that capture a specific pattern or solution. They live in ~/.claude/skills/ and are referenced when similar problems arise. The extraction process is triggered by enforcement hooks (Chapter 6: The Enforcement System) and the results are stored alongside your Life Repository (Chapter 2: The Life Repository).

# Skill: [Pattern Name]

## Pattern
[What this skill solves]

## When This Applies
- [Trigger condition 1]
- [Trigger condition 2]

## Solution
[Step-by-step solution]

## Example
[Concrete example of the pattern in action]

## Source
Discovered on [date] while working on [context].

Knowledge Extraction Workflow

Knowledge does not capture itself. The Genesis Framework uses a deliberate extraction process: when an agent discovers something novel, it gets evaluated for reusability and captured as a skill or knowledge file.

1. 1Discovery: An agent finds something non-obvious while working (e.g., a workaround, a pattern, a configuration quirk).
2. 2Evaluation: Is it novel? Is it reusable across multiple situations? Is it non-obvious? Is it actionable? All four must be true.
3. 3Extraction: If it passes evaluation, create a skill file with the standard template.
4. 4Registration: Add the skill to knowledge_index.md so other agents can find it.
5. 5Verification: Confirm the skill file exists and is properly indexed.

The Self-Improvement Loop

The most powerful pattern in Genesis is not any single agent — it is the system's ability to improve itself. After every debugging session, mistake, or discovery, the system asks: 'Should this be captured to prevent recurrence?'

This creates a compounding knowledge effect. Each mistake makes the system smarter. Each discovery becomes a reusable skill.

1. 1Criteria for extraction: Would this save more than 15 minutes if known upfront? Does it apply to more than one scenario? Was the solution non-obvious?
2. 24 quality gates: Novelty (not already documented), Reusability (applies broadly), Non-Obvious (required investigation), Actionable (clear steps to follow).
3. 3Skill file format: Title, Pattern, When This Applies, Solution, Example.
4. 4Registration: Add the skill to your knowledge index so other agents can find it.
5. 5Trigger: After any debugging session over 5 minutes, automatically evaluate if the solution should be extracted.

The knowledge base feeds directly into agents (Chapter 3: Building Your First Agents) by giving them domain expertise on demand. When an agent faces a familiar problem, it checks the knowledge index before starting fresh research. The autonomous loop in Chapter 11 also produces results logs that become knowledge entries.