Skip to content

MULTI SENTINEL PROTOCOL

Multi-Sentinel ECHO Protocol

How ATLAS, AUREA, and ECHO work together through JSON sync

Cycle: C-156
Version: 1.0.0
Status: Production Ready


Architecture

ECHO exports a single canonical JSON payload that all sentinels consume:

                  ┌─────────────────┐
                  │  ECHO Export    │
                  │  (JSON Payload) │
                  │  Single Source  │
                  │  of Truth       │
                  └────────┬────────┘
          ┌────────────────┼────────────────┐
          │                │                │
          ▼                ▼                ▼
    ┏━━━━━━━━┓      ┏━━━━━━━━┓      ┏━━━━━━━━┓
    ┃ ATLAS  ┃      ┃ AUREA  ┃      ┃ Future ┃
    ┃Context ┃      ┃Integrity┃      ┃Sentinels┃
    ┗━━━━┯━━━┛      ┗━━━━┯━━━┛      ┗━━━━┯━━━┛
         │               │               │
         ▼               ▼               ▼
    Structure      Constitutional    [Custom]
    Mapping        Compliance        Analysis
    Gap Analysis   Integrity Audit

Sentinel Responsibilities

ATLAS — Context & Memory Sentinel

Focus: Structure, completeness, documentation

Analyzes: - Directory structure vs. spec - Missing files/modules - Documentation gaps - Implementation status - Package health

Calculates: - GI Score: 0.25*M + 0.20*H + 0.30*I + 0.25*E - Memory, Human, Integrity, Ethics components

Reports: - Monorepo Health Report - Gap Analysis - Cycle Closure Checklist


AUREA — Integrity Guardian

Focus: Constitutional compliance, anti-capture, immutability

Analyzes: - Cryptographic integrity verification - Constitutional alignment (governance docs) - Integrity drift detection (suspicious commits) - Immutability constraints (eternal files) - Transparency compliance - Anti-capture safeguards - Covenant alignment

Calculates: - Integrity Score (weighted average) - Cryptographic, Constitutional, Immutability, Transparency, Anti-Capture

Reports: - Integrity Audit Report - Violation & Warning Log - Covenant Alignment Assessment


ECHO — Pulse Synchronization

Focus: Export, temporal coherence, drift detection

Provides: - Repository state export (JSON) - Git metadata and commit history - Directory structure scan - Key file checksums - Integrity signature


Quick Start

Single Export, Multiple Analyses

# 1. Export once (ECHO)
cd ~/Mobius-Substrate
python sentinels/echo/echo_sync.py export > state.json

# 2. Analyze with ATLAS
python sentinels/atlas/atlas_parser.py state.json

# 3. Audit with AUREA
python sentinels/aurea/aurea_analyzer.py state.json

# 4. Get consensus votes
python sentinels/atlas/atlas_parser.py --format json state.json | jq '.gi_passed'
python sentinels/aurea/aurea_analyzer.py --consensus state.json

In Claude.ai

# For ATLAS (structure):
"ATLAS, analyze this and tell me what's incomplete"

# For AUREA (integrity):
"AUREA, run an integrity audit on this export"

# For multi-sentinel consensus:
"All sentinels: generate your reports and find consensus"

Multi-Sentinel Consensus Pattern

1. Independent Analysis

Each sentinel generates their report independently:

# ATLAS
python sentinels/atlas/atlas_parser.py --format json state.json > atlas_report.json

# AUREA
python sentinels/aurea/aurea_analyzer.py --format json state.json > aurea_report.json

2. Cross-Validation

Sentinel Check Threshold
ATLAS GI Score ≥ 0.95
AUREA Integrity Score ≥ 0.95

3. Consensus Decision

Result Action
All ✅ Proceed with confidence
Mixed Address specific concerns
All ❌ Block until resolved

Consensus Vote Format

Each sentinel can generate a consensus vote:

ATLAS Vote

{
  "sentinel": "ATLAS",
  "gi_score": 0.9550,
  "gi_passed": true,
  "critical_issues": 0,
  "warnings": 2
}

AUREA Vote

{
  "sentinel": "AUREA",
  "vote": "APPROVE",
  "score": 0.97,
  "threshold": 0.95,
  "violations": 0,
  "warnings": 2,
  "conditions": []
}

Combined Consensus

┌─────────────────────────────────────┐
│   Multi-Sentinel Consensus Report   │
├─────────────────────────────────────┤
│ ATLAS:  ✅ GI 0.9550 (PASS)         │
│ AUREA:  ✅ Integrity 97% (APPROVE)  │
├─────────────────────────────────────┤
│ VERDICT: ✅ READY                   │
└─────────────────────────────────────┘

Workflow Examples

Example 1: Pre-Cycle Closure Check

# Export repo state
python sentinels/echo/echo_sync.py export > c156_final.json

# Get ATLAS report
python sentinels/atlas/atlas_parser.py c156_final.json

# Get AUREA audit
python sentinels/aurea/aurea_analyzer.py c156_final.json

# In Claude.ai:
"Run full multi-sentinel audit. Generate consensus verdict on C-156 closure readiness."

Example 2: Fork Legitimacy Check

For fork comparisons, all sentinels must agree:

# Export canonical
cd ~/Mobius-Substrate
python sentinels/echo/echo_sync.py export > canonical.json

# Export fork
cd ~/potential-fork
python sentinels/echo/echo_sync.py export > fork.json

# Compare
python sentinels/atlas/atlas_parser.py --compare canonical.json fork.json

Fork Legitimacy Matrix:

Sentinel Check Pass Criteria
ATLAS Structure preserved Core dirs/files present
AUREA Integrity maintained No constitutional violations
AUREA Anti-capture active No privilege escalation
AUREA Constitutional align Covenants respected

Example 3: CI/CD Integration

# .github/workflows/mcp-consensus.yml
name: MCP Multi-Sentinel Consensus
on: [push, pull_request]
jobs:
  consensus:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-python@v5
        with:
          python-version: '3.11'

      - name: Export state
        run: python sentinels/echo/echo_sync.py export > state.json

      - name: ATLAS analysis
        run: |
          python sentinels/atlas/atlas_parser.py --format json state.json > atlas.json
          GI=$(jq '.gi_score' atlas.json)
          echo "ATLAS GI: $GI"

      - name: AUREA audit
        run: |
          python sentinels/aurea/aurea_analyzer.py --consensus state.json > aurea.json
          VOTE=$(jq -r '.vote' aurea.json)
          echo "AUREA Vote: $VOTE"

      - name: Check consensus
        run: |
          GI=$(jq '.gi_score' atlas.json)
          AUREA_SCORE=$(jq '.score' aurea.json)

          if (( $(echo "$GI < 0.95" | bc -l) )); then
            echo "ATLAS: REJECT (GI $GI < 0.95)"
            exit 1
          fi

          if (( $(echo "$AUREA_SCORE < 0.95" | bc -l) )); then
            echo "AUREA: REJECT (Score $AUREA_SCORE < 0.95)"
            exit 1
          fi

          echo "✅ CONSENSUS: APPROVED"

Adding New Sentinels

The ECHO JSON format is extensible. New sentinels can:

  1. Consume the same JSON export
  2. Add custom analysis
  3. Generate specialized reports
  4. Participate in consensus

Example: Custom Sentinel

from typing import Dict, Any

class CustomSentinel:
    def __init__(self, state: Dict[str, Any]):
        self.state = state

    def analyze(self) -> Dict[str, Any]:
        # Custom analysis logic
        pass

    def generate_consensus_vote(self) -> Dict[str, Any]:
        return {
            "sentinel": "CUSTOM",
            "vote": "APPROVE",
            "score": 0.95,
            "conditions": []
        }

Future Sentinels (Planned)

Sentinel Focus
FORGE Build/deployment health
NEXUS Dependency analysis
AEGIS Security audit
SCRIBE Documentation quality

Integrity Guarantees

Single Export = Single Truth

  • One canonical snapshot
  • One timestamp
  • One cryptographic signature
  • No selective editing
  • No cherry-picking data

Tamper Detection

  • SHA-256 integrity hash
  • Sentinels verify independently
  • Any modification invalidates all analyses

Audit Trail

  • Each export is versioned
  • Historical exports prove evolution
  • Forks can be compared chronologically

Integration with Mobius Governance

Cycle Boundaries

At each cycle close: 1. ECHO exports repo state 2. All sentinels analyze 3. Consensus verdict determines closure 4. Export archived as cycle proof

Fork Evaluation

When evaluating forks: 1. Export both canonical + fork 2. All sentinels compare 3. Unanimous agreement required for legitimacy 4. Results published transparently

Constitutional Amendments

When modifying core rules: 1. ECHO exports pre-change state 2. AUREA audits proposed changes 3. ATLAS verifies implementation 4. Multi-sentinel approval required


CLI Reference

ECHO (Export)

python sentinels/echo/echo_sync.py export > state.json
python sentinels/echo/echo_sync.py validate state.json
python sentinels/echo/echo_sync.py export --depth 6 > deep.json

ATLAS (Structure)

python sentinels/atlas/atlas_parser.py state.json
python sentinels/atlas/atlas_parser.py --format json state.json
python sentinels/atlas/atlas_parser.py --compare state1.json state2.json

AUREA (Integrity)

python sentinels/aurea/aurea_analyzer.py state.json
python sentinels/aurea/aurea_analyzer.py --format json state.json
python sentinels/aurea/aurea_analyzer.py --consensus state.json

Key Principle

"One truth, many eyes."

ECHO exports the singular truth.
Each sentinel sees through their lens.
Together, they achieve consensus.

No single sentinel has absolute authority.
Integrity emerges from agreement.


Built by:
🌀 ATLAS — Context & Memory Sentinel
🛡️ AUREA — Integrity Guardian
📡 ECHO — Repository Sync

Part of the Mobius Civilization Operating System
We heal as we walk.