Evaluate: Should /debug command have dedicated sub-agent?

[dean0x]

Context

Currently /debug is a 227-line inline command without a dedicated sub-agent. Other commands like /catch-up (29 lines) and /commit (28 lines) delegate to sub-agents that handle the heavy lifting.

Current State

Command: src/claude/commands/devflow/debug.md (227 lines)

• Full inline workflow implementation
• Uses: Bash, Read, Write, Edit, Grep, Glob, TodoWrite
• Handles systematic debugging with issue tracking

No Agent: No corresponding src/claude/agents/devflow/debug.md

Question

Should we create a dedicated debug sub-agent to:

1. Reduce token usage in command execution
2. Separate concerns (command = UX, agent = implementation)
3. Follow pattern used by catch-up, commit, research commands

Trade-offs

Pros of Creating Agent:

• Consistent pattern with other commands
• Lower token cost per invocation
• Easier to maintain debugging logic separately
• Agent could be reused outside slash command context

Cons of Creating Agent:

• Debug workflow may need interactive user input (like /release)
• Additional file/complexity to maintain
• Current inline approach works fine

Decision Needed

Should /debug follow the simple wrapper pattern (launch agent) or keep inline workflow?

Related

• Evaluate: Should /devlog command have dedicated sub-agent? #12 (devlog agent evaluation)
• Evaluate: Should /plan-next-steps command have dedicated sub-agent? #13 (plan-next-steps agent evaluation)

[dean0x]

✅ Resolved in v0.4.0

This was addressed as part of the Skills Infrastructure release with a dual-mode pattern.

Implementation

Created both a debug sub-agent AND a debug skill:

Agent: src/claude/agents/devflow/debug.md (12,079 bytes)

• Specialized agent for systematic debugging with hypothesis testing
• Tools: Bash, Read, Write, Edit, Grep, Glob, TodoWrite
• Used by /debug command for explicit user invocation

Skill: src/claude/skills/devflow/debug/

• Auto-activating skill that triggers on errors/failures
• Same debugging logic, different activation pattern
• Proactively helps when issues are detected

Command: /debug delegates to debug agent

• User controls when to invoke (manual mode)
• Follows orchestrator pattern like other commands

Dual-Mode Pattern

The debug functionality exists in two forms:

1. Command mode (/debug) - Manual, explicit user invocation
2. Skill mode (auto) - Automatic activation when errors detected

This gives users flexibility: manual control when preferred, automatic assistance when helpful.

Changes

• PR Add Skills Support and Command→Agent→Skill Architecture #15: Add Skills Support and Command→Agent→Skill Architecture
• Release: v0.4.0 (2025-10-21)

CHANGELOG Entry (v0.4.0)

#### Skills Infrastructure
- 7 new skills including debug skill for automatic activation
- Auto-activation based on conversation context

#### Command→Agent→Skill Architecture
- Dual-mode pattern for debug (command + skill)
- Commands for explicit invocation
- Skills for auto-activation

Benefits Achieved

✅ Reduced token usage - Command delegates to agent (12KB agent vs inline)
✅ Separation of concerns - Command = UX, Agent = implementation
✅ Consistent pattern - Follows catch-up, commit, research pattern
✅ Bonus: Skill mode - Auto-activation adds additional value beyond original request

Released: v0.4.0 (2025-10-21)
npm: https://www.npmjs.com/package/devflow-kit/v/0.4.0