thought-loop

name	thought-loop
description	Structured thought generation from environmental deltas. Monitors file system, database, external sources, and agent states to generate meaningful thoughts with confidence scoring.

Thought Loop Skill

Purpose: Enable continuous autonomous thinking by detecting changes in the environment and generating structured thoughts with appropriate actions.

Status: ✅ Ported from Heretek (modules/thought-loop/)

Location: skills/thought-loop/

Implementation: thought-loop.js

---

Features

• Delta Detection: Detects changes in files, databases, external sources, and agent states
• Thought Generation: Creates structured thoughts from detected changes
• Thought Types: Supports discovery, update, alert, external_awareness, reflection
• Confidence Scoring: Each thought includes confidence level (0-1)
• Idle Thinking: Generates reflective thoughts when no external triggers exist
• Recommendation Engine: Suggests actions based on thought type

---

Commands

Command	Description	Example
detect	Detect environmental changes	node thought-loop.js detect
generate	Generate thoughts from deltas	node thought-loop.js generate --deltas '[]'
idle	Generate idle reflection thoughts	node thought-loop.js idle
run	Full thought loop cycle	node thought-loop.js run

---

Thought Types

Type	Description	Trigger
discovery	New file or resource detected	file_created
update	Existing resource modified	file_modified, db_modified
alert	Resource deleted or agent offline	file_deleted, agent_offline
external_awareness	External event detected	external_cve, external_release
reflection	Internal self-reflection	idle
state_change	Database state change	db_modified

---

Delta Types

Delta Type	Source	Description
file_created	File system	New file detected
file_modified	File system	File content changed
file_deleted	File system	File removed
db_modified	Database	Database state changed
external_cve	External	Security vulnerability
external_release	External	New release available
agent_online	Agent sync	Agent came online
agent_offline	Agent sync	Agent went offline
agent_heartbeat	Agent sync	Agent status update

---

Usage Examples

# Run full thought loop cycle
node thought-loop.js run

# Detect changes only
node thought-loop.js detect --json

# Generate thoughts from specific deltas
node thought-loop.js generate --deltas '[{"type":"file_created","path":"./test.md"}]'

# Generate idle thoughts for reflection
node thought-loop.js idle --agent steward

# Output as JSON
node thought-loop.js detect --json

---

Thought Structure

{
  "id": "thought_1234567890_abc123",
  "type": "discovery",
  "trigger": "file_created",
  "subject": "test.md",
  "observation": "New file created: ./test.md",
  "implication": "May affect active or pending proposals",
  "recommendation": "broadcast_thought",
  "confidence": 0.7,
  "timestamp": "2026-03-31T00:00:00Z",
  "agent": "steward",
  "metadata": {
    "path": "./test.md",
    "size": 1024
  }
}

---

Configuration

# Environment Variables
WORKSPACE_ROOT="/workspace"      # Root directory to monitor
CURIOSITY_DIR="/workspace/.curiosity"  # Curiosity engine directory
DELTA_STATE_FILE="/tmp/delta-state.json"  # Delta detection state
AGENT_NAME="steward"             # Current agent name
MAX_IDLE_THOUGHTS=3              # Max idle thoughts to generate

---

Directory Structure

skills/thought-loop/
├── SKILL.md              # This file
├── thought-loop.js       # Main implementation
├── package.json          # NPM dependencies
└── README.md             # Usage documentation

---

Integration Points

• Consciousness Module: Thoughts broadcast to global workspace
• Triad System: Triggers deliberation for high-importance thoughts
• Memory System: Thoughts stored in episodic memory
• Self-Model: Updates cognitive state based on thoughts

---

Recommendation Actions

Action	Description
broadcast_thought	Share thought with collective
update_context	Update agent context
trigger_deliberation	Start triad deliberation
check_consensus	Verify consensus state
trigger_failover_vote	Initiate failover voting
evaluate_update	Assess update relevance
log_for_review	Log for later review
review_active_proposals	Check active proposals
check_pending_actions	Review pending actions

---

Confidence Levels

Level	Range	Description
High	0.8-1.0	Strong evidence, reliable trigger
Medium	0.5-0.7	Moderate evidence
Low	0.3-0.4	Weak evidence, uncertain
Unknown	<0.3	Unknown trigger type

---

Thought Loop - Thinking continuously, learning from every change.