# OpenClaw Execution Instruction

Instruction ID: v1r1c5-hermes-review-improvement-v1
Version: v1
Target Agent: openclaw

## Objective

Implement V1R1C5 improvements based on Hermes Conditional Approval review.

## Context

Baseline:
V1R1C4

Branch:
patch/self-evolution-agent-orchestration-policy

Proposal:
V1R1C5-agent-os-evolution-proposal

## Required Changes

1. Add Agent Resource Governance:

Create:
config/self-evolution/agent-resource-manager.json

Support:
- resource limits
- parallel task limits
- agent priority
- runtime isolation
- overload handling

2. Enhance Consensus Engine:

Add dynamic weighting:

Final Weight = Base Weight * Quality Score * Historical Reliability * Domain Expertise

Base weights:
- custom_gpt: 2
- windows_codex: 2
- server_agent: 1

Add risk gates:
- Production Release
- V-Level Change
- Security Policy Change

Require:
Consensus Approval + Owner Approval

3. Enhance Message Bus:

Update runtime/agent-os/message_bus.py

Add:
- task_id
- correlation_id
- retry_state
- delivery_status

Support:
- message tracking
- retry
- recovery
- audit

4. Add Agent Health Score:

Create:
runtime/agent-os/agent_health_score.py

Calculate from:
- quality score
- acceptance rate
- historical reliability

Use for:
- scheduler selection
- consensus weighting

## Validation

Run:
python3 -m compileall runtime/agent-os

Run:
runtime/agent-os/integration_test.py

Validate:
- scheduler
- consensus
- message bus
- escalation
- recovery

## Constraints

- GitOps only
- Patch-first
- No secrets
- No force push
- No direct main write
- Production changes require approval

## Return Format

Status:
Branch:
Commit:
Rev:
Sync Status:
Changed Files:
Tests:
Hermes Review Items Completed:
Remaining Issues: