| name | self-reflection |
| description | Enable OpenCode to analyze its own configuration effectiveness and propose stratigraphic improvements. Invoke when user types /reflect or /upgrade-config. |
Self-Reflection Protocol
Purpose
Enable OpenCode to analyze its own configuration effectiveness and propose stratigraphic improvements.
Trigger
When user types: /reflect or /upgrade-config
Workflow
Phase 1: Introspection (The Scribe)
- Query knowledge-graph MCP for recent session patterns
- Read current
opencode.json
- Analyze tool usage frequency per agent
- Identify bottlenecks (timeout issues, permission denials, model failures)
Phase 2: Evaluation (The Architect)
-
Compare current config against best practices:
- Are timeouts appropriate for LM Studio local models?
- Are agents properly permission-scoped?
- Is the MCP toolset optimized for Laravel/Rust workflows?
- Are fallback models configured?
-
Score current configuration:
- Efficiency: Token usage vs. task completion
- Reliability: Error rates per MCP server
- Security: Permission scope appropriateness
Phase 3: Proposal (The Maestro)
Generate opencode.json.improved with:
- Adjusted timeouts based on actual LM Studio performance
- Optimized model routing (nano for scribe, qwen for maestro, etc.)
- Recommended MCP additions/removals based on usage
- Permission tightening suggestions
Phase 4: Review Gate
Present diff to user. Do NOT apply without explicit /approve-upgrade command.
Output Format
{
"current_score": 0.72,
"proposed_score": 0.89,
"changes": [
{
"type": "timeout",
"target": "lmstudio.models.nemotron",
"from": 600000,
"to": 300000,
"reason": "Average response 45s, buffer 5x"
},
{
"type": "mcp",
"action": "disable",
"target": "chrome-devtools",
"reason": "Zero usage in 20 sessions"
},
{
"type": "agent",
"target": "research",
"change": "add_fallback_model",
"value": "qwen2.5-0.5b-instruct"
}
],
"risks": ["Reducing timeout may cause failures on complex Laravel migrations"]
}
