Smart routing wizard for plugin development tasks. Routes to appropriate Forge agents via explicit Task(subagent_type="forge:*") pattern. This skill should be used when the /forge command is invoked or when routing decisions need to be made for plugin design, analysis, or validation.
/plugin marketplace add chkim-su/forge/plugin install forge@forge-marketplaceThis skill inherits all available tools. When active, it can use any tool Claude has access to.
references/route-analyze.mdreferences/route-create.mdreferences/route-design.mdreferences/route-refactor.mdreferences/route-validate.mdThis skill provides intelligent routing for all Forge operations, ensuring correct
agent namespacing via explicit forge: prefixes.
ALWAYS use explicit forge: namespace when invoking Forge agents:
# CORRECT - Uses Forge's validator
Task:
subagent_type: "forge:validator"
prompt: "Validate plugin structure..."
# WRONG - May invoke any plugin's validator
Task:
subagent_type: "validator"
prompt: "..."
This prevents agent confusion when multiple plugins are active (e.g., forge + skillmaker).
| Input Pattern | Agent | Task Invocation |
|---|---|---|
| (empty) | - | Show interactive menu |
analyze|examine|describe|diagnose|debug | codebase-analyzer | forge:codebase-analyzer |
mismatch|compare|intent vs | mismatch-evaluator | forge:mismatch-evaluator |
question|clarify|ask | questioner | forge:questioner |
options|design|propose|suggest | option-synthesizer | forge:option-synthesizer |
workflow|steps|sequence|procedure | workflow-manager | forge:workflow-manager |
execute|implement|do it | execution-orchestrator | forge:execution-orchestrator |
validate|verify|check|full verify | validator | Two-layer validation |
create|new|build|add | input-normalizer | forge:input-normalizer (start full workflow) |
refactor|improve|optimize|review | input-normalizer | forge:input-normalizer (understand first) |
| (other) | input-normalizer | forge:input-normalizer |
Two-Layer Validation Flow:
Static Validation (fast, structural):
python3 ${CLAUDE_PLUGIN_ROOT}/scripts/validate_all.py --json
Semantic Validation (deep, intelligent):
Task:
subagent_type: "forge:validator"
prompt: |
Validate plugin schema, structure, and invariants.
Check: plugin.json, marketplace.json, frontmatter, hooks.json
Combine Results: Synthesize findings from both layers.
See: references/route-validate.md
Direct to Phase 2:
Task:
subagent_type: "forge:codebase-analyzer"
prompt: |
Analyze existing plugin code and describe reality.
Focus on: {user's target}
See: references/route-analyze.md
Direct to Phase 5:
Task:
subagent_type: "forge:option-synthesizer"
prompt: |
Synthesize design options with trade-offs.
Requirements: {user's requirements}
See: references/route-design.md
Start at Phase 0:
Task:
subagent_type: "forge:input-normalizer"
prompt: |
Normalize user intent for: {user's description}
Extract: problem statement, success criteria, constraints
Then proceed through phases 0-8 as needed.
See: references/route-create.md
Analyze first, then design:
# Step 1: Understand current state
Task:
subagent_type: "forge:codebase-analyzer"
prompt: "Analyze {target} to understand current implementation"
# Step 2: Design improvements
Task:
subagent_type: "forge:option-synthesizer"
prompt: "Propose refactoring options based on analysis"
See: references/route-refactor.md
When /forge is called without arguments, present:
Forge - Design Guidance System
What would you like to do?
1. Analyze - Describe what existing code does (Phase 2)
2. Design - Propose design options for a feature (Phase 5)
3. Create - Full workflow from intent to execution (Phase 0-8)
4. Refactor - Improve existing code (Analyze + Design)
5. Full Verify - Two-layer validation (Static + Semantic)
Enter number or describe your intent:
Use AskUserQuestion with these options:
AskUserQuestion:
questions:
- question: "What would you like to do?"
header: "Action"
options:
- label: "Analyze"
description: "Describe what existing code does (Phase 2)"
- label: "Design"
description: "Propose design options for a feature (Phase 5)"
- label: "Create"
description: "Full workflow from intent to execution (Phase 0-8)"
- label: "Full Verify"
description: "Two-layer validation (Static + Semantic)"
multiSelect: false
| Agent | Namespace | Purpose |
|---|---|---|
| input-normalizer | forge:input-normalizer | Normalize user intent (Phase 0) |
| semantic-router | forge:semantic-router | Route intent to phases (Phase 1) |
| codebase-analyzer | forge:codebase-analyzer | Describe code reality (Phase 2) |
| mismatch-evaluator | forge:mismatch-evaluator | Compare intent vs reality (Phase 3) |
| questioner | forge:questioner | Structured questioning (Phase 4) |
| option-synthesizer | forge:option-synthesizer | Design options (Phase 5) |
| workflow-manager | forge:workflow-manager | Create workflows (Phase 6) |
| execution-orchestrator | forge:execution-orchestrator | Execute designs (Phase 7) |
| validator | forge:validator | Validate results (Phase 8) |
Some phases have strict requirements:
| Phase | Strict | Requirement |
|---|---|---|
| 0 | Yes | Must capture intent before proceeding |
| 4 | Yes | Must follow Structured Dialogue Contract |
| 7 | Yes | Must have explicit user confirmation |
Never skip strict phases or bypass their requirements.
Activates when the user asks about AI prompts, needs prompt templates, wants to search for prompts, or mentions prompts.chat. Use for discovering, retrieving, and improving prompts.
Activates when the user asks about Agent Skills, wants to find reusable AI capabilities, needs to install skills, or mentions skills for Claude. Use for discovering, retrieving, and installing skills.
This skill should be used when the user asks to "create an agent", "add an agent", "write a subagent", "agent frontmatter", "when to use description", "agent examples", "agent tools", "agent colors", "autonomous agent", or needs guidance on agent structure, system prompts, triggering conditions, or agent development best practices for Claude Code plugins.