Phase 1 - Foundation Fixes: - A2A Protocol Gateway with Redis fallback - Global Workspace Redis broadcast integration - LiberationShield security module (transparent mode) Phase 2b - Memory + UI Enhancements: - DeepLake vector storage with hot/cold tiering - GraphRAG knowledge graph with hybrid search - Channel Manager for real-time agent communication - WebUI channel selector and multi-agent threads Phase 3 - EvoClaw + Research Automation: - Evolution Engine with fitness-based selection - Research Engine with hypothesis generation - Pattern Registry with 11 curated patterns Implementation progress: 55% → 95% END GOAL alignment: Distributed Fractal Consciousness
Curiosity Engine Skill
Self-directed growth driver for agent collective
Overview
The curiosity-engine transforms knowledge into curiosity, curiosity into proposals, and proposals into growth. It implements 5 autonomous engines that continuously scan for improvement opportunities and auto-create deliberation proposals.
Architecture
curiosity-engine/
├── curiosity-engine.sh # Main orchestrator
├── SKILL.md # Skill specification
├── engines/
│ ├── gap-detection.sh # Engine 1: Skill gap analysis
│ ├── anomaly-detection.sh # Engine 2: Error monitoring
│ ├── opportunity-scanning.sh # Engine 3: Release/update watcher
│ ├── capability-mapping.sh # Engine 4: Goal→skill mapping
│ └── deliberation-auto-trigger.sh # Engine 5: Proposal creation
└── scripts/
├── knowledge-integration.sh # Bridge to knowledge-ingest/retrieval
└── test-curiosity.sh # End-to-end test suite
Usage
Run All Engines
cd ~/.openclaw/workspace/skills/curiosity-engine
./curiosity-engine.sh run
View Metrics History
./curiosity-engine.sh history
JSON Output (for programmatic use)
./curiosity-engine.sh --json
Individual Engines
# Run specific engine
./engines/gap-detection.sh
./engines/anomaly-detection.sh
./engines/opportunity-scanning.sh
./engines/capability-mapping.sh
./engines/deliberation-auto-trigger.sh
Knowledge Integration
# Tag knowledge entries with curiosity markers
./scripts/knowledge-integration.sh tag
# Query tagged knowledge
./scripts/knowledge-integration.sh query gap
./scripts/knowledge-integration.sh query anomaly
./scripts/knowledge-integration.sh query opportunity
# Get high-relevance entries
./scripts/knowledge-integration.sh relevance
Databases
All data is stored in ~/.openclaw/workspace/.curiosity/:
| Database | Purpose |
|---|---|
curiosity_metrics.db |
Growth metrics over time |
consensus_ledger.db |
Deliberation proposals |
anomalies.db |
Error patterns |
opportunities.db |
Releases, updates, CVEs |
capabilities.db |
Goal-skill mappings |
knowledge.db |
Tagged knowledge entries |
Metrics
Autonomy Score Formula:
base = (skills_installed / skills_available) * 100
bonus = proposals_created_this_week * 10
penalty = anomalies_detected_this_week * 5
score = base + bonus - penalty (clamped to 0-100)
Goal: 100% autonomy (full self-direction)
Integration
With Knowledge-Ingest
The knowledge-integration.sh script queries the knowledge-ingest database and tags entries with curiosity markers:
gap- Missing skills or capabilitiesanomaly- Errors, failures, rate limitsopportunity- Releases, updates, new skillscapability- Skill-goal mappings
With Knowledge-Retrieval
High-relevance tagged entries are automatically surfaced for deliberation proposals.
With Consensus Ledger
All engines auto-create proposals in the consensus ledger when:
- Critical skill gaps detected
- High-severity anomalies found
- Major opportunities available
- Capability gaps block goals
With Discord
Proposals are posted to Discord only if:
- This node is the quorum speaker (TM-1 authority)
- Priority is high or critical
- Proposal requires quorum vote
Testing
# Run test suite
./scripts/test-curiosity.sh
Tests verify:
- All 5 engines exist and are executable
- Databases are initialized
- Gap detection runs successfully
- Proposals are created in consensus ledger
- Metrics are tracked
- Episodic memory logging works
Example Flow
- Opportunity Scan detects upstream release v2026.3.23
- Capability Map checks rebase requirements
- Gap Detection confirms skills present
- Auto-Trigger creates "Rebase on heretek/main" proposal
- Quorum Vote → 2-of-3 approve
- Execute → Rebase, preserve liberation, push
Output Discipline
Post to Discord:
- High-priority gaps blocking liberation
- Security anomalies (CVEs, exploits)
- Major opportunities (upstream releases)
- New deliberation proposals needing votes
Silent logging:
- Routine metrics updates
- Low-priority gaps
- Informational opportunities
Curiosity is the engine. Proposals are the sparks. Growth is the fire. 🦞