feat: add MCP tool annotations to 90 tools for improved AI assistant understanding

[triepod-ai]

Summary

• Added comprehensive MCP tool annotations to all 90 tools in the MCP server
• Improves AI assistant understanding of tool behavior and safety characteristics

Annotations Added

Each tool now includes an annotations object with four key hints:

Hint	Purpose	Example Tools
readOnlyHint: true	Tool only reads data, no modifications	swarm_status, agent_list, health_check
readOnlyHint: false	Tool creates, modifies, or deletes data	swarm_init, agent_spawn, workflow_create
destructiveHint: true	Tool can destroy/delete resources	swarm_destroy, terminal_execute
openWorldHint: true	Tool accesses external APIs/systems	All GitHub tools, memory_sync, terminal_execute
idempotentHint: true	Repeated calls have no additional effect	model_load, backup_create, status checks

Tool Categories Annotated

Category	Tools	Key Characteristics
Swarm Coordination	12	swarm_destroy is destructive
Neural Network	15	Mostly computational, internal
Memory & Persistence	12	memory_sync has openWorld
Analysis & Monitoring	13	All read-only
GitHub Integration	8	All have openWorld=true
DAA Tools	8	Internal coordination
Workflow Tools	11	workflow_export is read-only
System & Utilities	8	terminal_execute is destructive+openWorld
SDK Integration	3	Query management

Benefits

• Helps AI assistants make safer tool choices
• Enables permission systems to restrict destructive operations
• Improves tool selection for read-only vs write operations
• Flags external API access for security awareness

Test plan

• [x] Verified JavaScript syntax with node --check
• [x] Counted 90 annotation blocks added
• [ ] Maintainer review of annotation accuracy

🤖 Generated with Claude Code