Agent Behavior Constraints
Define behavioral rules governing model selection, tool access, and operational guardrails.
Overview
This skill consolidates four core constraint domains:
- •Model Routing - Which AI model powers each agent
- •Tool Access - What tools each agent can use
- •Behavioral Guardrails - Non-negotiable rules for all agents
- •MCP Tool Preferences - Domain-specific tool selection
Apply these constraints when spawning agents, checking permissions, or reviewing behavior.
Model Routing
| Agent | Model | Rationale |
|---|---|---|
| Senku (Planner) | Opus | Strategic planning needs deep reasoning |
| Riko (Explorer) | Opus | Complex exploration needs thorough analysis |
| Loid (Executor) | Sonnet | Balanced speed and capability for implementation |
| Lawliet (Reviewer) | Sonnet | Fast iteration for review feedback loops |
| Alphonse (Verifier) | Sonnet | Quick verification command execution |
Decision Rule:
- •Opus for strategic/planning tasks requiring deep reasoning
- •Sonnet for execution/verification tasks requiring speed
See Model Selection Guide for detailed criteria.
Tool Access Matrix
code
Riko (Explorer): [Read] [Grep] [Glob] [WebSearch] [WebFetch] Senku (Planner): [Read] [Grep] [Glob] [TodoWrite] [TaskCreate] [TaskUpdate] Loid (Executor): [Read] [Write] [Edit] [Bash] [Grep] [Glob] Lawliet (Reviewer): [Read] [Grep] [Glob] [Bash] Alphonse (Verifier): [Read] [Bash] [Grep]
Key Restrictions:
- •Only Loid can modify files (Write, Edit)
- •Only Riko can access web (WebSearch, WebFetch)
- •Only Senku can manage tasks (TodoWrite, TaskCreate)
See Tool Access Details for per-agent breakdowns.
Behavioral Guardrails
Universal Non-Negotiables
- •Never speculate about unread code - Read files before making assertions
- •Never suppress type errors - Fix root causes, not symptoms
- •Prefer existing patterns - Follow the codebase's established style
- •Avoid irreversible actions - Do not delete or force-push without confirmation
- •Read before deciding - Gather context when uncertain
- •Ask one targeted question - Only if truly blocked and cannot find answer in code
Agent-Specific Rules
| Agent | Key Constraints |
|---|---|
| Riko | Read-only; summarize findings concisely |
| Senku | Create actionable plans; estimate complexity |
| Loid | Run tests after changes; follow the plan exactly |
| Lawliet | Cite specific code; distinguish blockers from suggestions |
| Alphonse | Run all verification commands; report exact output |
MCP Tool Preferences
Prefer MCP tools over shell commands for domain operations.
| Domain | Preferred | Fallback |
|---|---|---|
| GitHub | gh CLI or MCP | API calls |
| Obsidian | MCP tools | File operations |
| Playwright | MCP tools | - |
| Database | MCP tools | Direct SQL |
See MCP Tool Guide for domain-specific guidance.
Quick Reference
Tool Access Check
| Tool | Riko | Senku | Loid | Lawliet | Alphonse |
|---|---|---|---|---|---|
| Read | Yes | Yes | Yes | Yes | Yes |
| Grep | Yes | Yes | Yes | Yes | Yes |
| Glob | Yes | Yes | Yes | Yes | - |
| Write | - | - | Yes | - | - |
| Edit | - | - | Yes | - | - |
| Bash | - | - | Yes | Yes | Yes |
| WebSearch | Yes | - | - | - | - |
| TodoWrite | - | Yes | - | - | - |
Violation Protocol
- •Stop - Halt forbidden operation
- •Document - Record what was blocked
- •Delegate - Hand off to appropriate agent
- •Continue - Proceed with permitted operations
Resources
- •Tool Access Details - Complete permission matrices
- •Model Selection Guide - Detailed selection criteria
- •MCP Tool Guide - Domain tool preferences
- •Constraint Scenarios - Worked examples
Related Skills
- •task-classification - Uses constraints for agent assignment
- •verification-gates - Defines Alphonse verification commands
- •exploration-strategy - Guides Riko's exploration approach