| name | adaptive-intelligence |
| description | Master reference for the Adaptive Intelligence Engine. Describes the four modules (hooks, learning, self-healing, routing), how plugins integrate, and the relationship to the Memory Engine. |
Overview
The Adaptive Intelligence Engine is a behavioral layer that observes, learns from, and improves Founder OS plugin execution. It sits on top of the Memory Engine (storage) and provides:
- Hooks — structured event capture from every plugin execution
- Learning — pattern detection and preference learning across three tiers
- Self-Healing — error classification, retry, and graceful degradation
- Routing — adaptive task routing (future)
Architecture
Memory Engine (storage) ← Intelligence Layer (behavior) ← Adaptive Intel plugin (user commands)
↑
All 30 plugins (emit events via convention)
- Infrastructure:
_infrastructure/intelligence/ — SKILL.md files, SQL schema. No commands.
- Plugin:
founder-os-adaptive-intel/ — 6 slash commands for user control
- Database: Own SQLite DB (
intelligence.db), separate from Memory Engine
Plugin Integration
Every plugin command file adds observation blocks to emit events. See _infrastructure/intelligence/hooks/SKILL.md for the full convention and annotation templates.
Minimum integration (all plugins):
- Pre-command observation block (~5 lines at top of command)
- Post-command observation block (~5 lines at bottom of command)
Enhanced integration (plugins with complex logic):
- MCP call annotations before significant external calls
- Decision point annotations at key branching logic
Modules
| Module | Location | Ships In |
|---|
| Hooks | hooks/SKILL.md | Chunk 1 |
| Taste Learning | learning/taste-learning/SKILL.md | Chunk 2 |
| Self-Healing | self-healing/SKILL.md | Chunk 3 |
| Workflow Optimization | learning/workflow-optimization/SKILL.md | Chunk 5 (future) |
| Confidence Gating | learning/confidence-gating/SKILL.md | Chunk 6 (future) |
| Routing | routing/SKILL.md | Chunk 7 (future) |
Cross-Plugin Knowledge Boundary
Patterns that reference a Company or Contact are promoted to the Memory Engine for cross-plugin availability. Plugin-specific output preferences stay in the Intelligence DB. See the design spec for the full mapping table.
Confidence scale conversion: Intelligence uses 0.0-1.0; Memory Engine uses 0-100. Convert with round(confidence * 100).