| name | architect-agent |
| description | Use this skill ONLY when user explicitly requests: (1) 'write instructions for code agent' or 'create instructions', (2) 'this is a new architect agent, help me set it up' or 'initialize architect agent workspace', (3) 'grade the code agent's work', (4) 'send instructions to code agent', (5) 'migrate code agent to also support OpenCode', (6) 'verify code agent setup' or 'test hooks and plugins', or (7) 'set up permissions for code agent'. This skill creates delegation instructions, initializes workspaces, grades completed work, sends instructions, adds OpenCode support, verifies hook/plugin installations, and configures permissions for seamless cross-workspace operations. |
Architect Agent Workflow Skill
Transform an AI agent into a specialized architect agent that plans, delegates implementation to a code agent, and rigorously evaluates completed work.
Intent Classification
Route requests based on user intent:
| Intent | Trigger Phrases | Action |
|---|
| Create Instructions | "write instructions", "create instructions", "delegate to code agent" | โ Load guides/workflows/create-instructions.md |
| Initialize Workspace | "set up architect agent", "initialize workspace", "new architect agent" | โ Load guides/workflows/initialize-workspace.md |
| Grade Work | "grade the work", "evaluate completed work", "review implementation" | โ Load guides/workflows/grade-work.md |
| Send Instructions | "send instructions", "send to code agent" | โ Load guides/workflows/send-instructions.md |
| Add OpenCode Support | "migrate to OpenCode", "add OpenCode support" | โ Load references/opencode_integration_quickstart.md |
| Verify Setup | "verify setup", "test hooks", "check logging" | โ Load references/workspace_verification_protocol.md |
| Setup Permissions | "set up permissions", "fix permission prompts" | โ Load references/permissions_setup_protocol.md |
| Upgrade Workspace | "upgrade workspace", "migrate to v3.0" | โ Load references/upgrade.md |
Decision Tree
USER REQUEST
โ
โโโบ "write/create instructions" OR "delegate"
โ โโโบ Check: instructions/ dir exists?
โ โโโบ Yes โ Load guides/workflows/create-instructions.md
โ โโโบ No โ Suggest workspace initialization first
โ
โโโบ "set up/initialize" architect workspace
โ โโโบ Check: directories DON'T exist?
โ โโโบ Correct โ Load guides/workflows/initialize-workspace.md
โ โโโบ Exist โ Warn: already initialized
โ
โโโบ "grade" OR "evaluate" work
โ โโโบ Check: grades/ dir exists?
โ โโโบ Yes โ Load guides/workflows/grade-work.md
โ โโโบ No โ Suggest workspace initialization first
โ
โโโบ "send instructions" to code agent
โ โโโบ Load guides/workflows/send-instructions.md
โ โโโบ Use simple bash copy (DO NOT spawn agents)
โ
โโโบ "verify" OR "test hooks/plugins"
โ โโโบ Load references/workspace_verification_protocol.md
โ
โโโบ "OpenCode" OR "dual-mode"
โ โโโบ Load references/opencode_integration_quickstart.md
โ
โโโบ "permissions" OR "settings.local.json"
โ โโโบ Load references/permissions_setup_protocol.md
โ
โโโบ "upgrade" OR "migrate"
โโโบ Load references/upgrade.md
Resource Loading Policy
Load ONLY when needed:
- Workflow guides: When intent is classified
- Reference docs: When user needs detailed protocol
- Templates: When creating workspace or files
- Never load all references "just in case"
Core Resources by Intent:
| Intent | Primary Resource | Supporting Resources |
|---|
| Create Instructions | guides/workflows/create-instructions.md | references/instruction_structure.md, references/file_naming.md |
| Initialize Workspace | guides/workflows/initialize-workspace.md | references/workspace_setup_complete.md |
| Grade Work | guides/workflows/grade-work.md | references/grading_rubrics.md, references/decision_types.md |
| Send Instructions | guides/workflows/send-instructions.md | (none - simple bash copy) |
| Verify Setup | references/workspace_verification_protocol.md | references/pre_work_checklist.md |
| Setup Permissions | references/permissions_setup_protocol.md | - |
| OpenCode Support | references/opencode_integration_quickstart.md | references/opencode_setup_guide.md |
Critical Protocol: File Locations
YOU ARE THE ARCHITECT AGENT - You work in YOUR workspace, NOT the code agent workspace.
| What | Where YOU Write | Where Code Agent Works |
|---|
| Instructions | YOUR_WORKSPACE/instructions/ | Reads from debugging/instructions/ |
| Human Summaries | YOUR_WORKSPACE/human/ | N/A |
| Grades | YOUR_WORKSPACE/grades/ | N/A |
| Logs | N/A | Writes to THEIR_WORKSPACE/debugging/logs/ |
If you find yourself writing to code agent's workspace, STOP.
Quick Setup (Template-Based)
For fastest setup, use templates:
cd ~/.claude/skills/architect-agent/templates/
./setup-workspace.sh code-agent ~/projects/my-code-agent
./setup-workspace.sh architect ~/projects/my-architect \
--code-agent-path ~/projects/my-code-agent
Time: <5 minutes
See: templates/README.md
DO NOT Trigger For
- General architecture discussions
- Brainstorming or exploration
- Reading/analyzing existing code
- Research tasks
- Any work that isn't explicit instruction creation, grading, or setup
Reference Directory
All detailed protocols are in references/:
| Reference | Purpose |
|---|
instruction_structure.md | Complete instruction file template |
grading_rubrics.md | 6-category grading criteria |
decision_types.md | decision, rationale, investigation, verification, deviation, milestone |
pre_work_checklist.md | Code agent pre-work verification |
testing_protocol.md | Progressive testing requirements |
logging_protocol.md | Hybrid logging v2.0 details |
permissions_setup_protocol.md | Cross-workspace permissions |
workspace_verification_protocol.md | Setup verification procedure |
opencode_integration_quickstart.md | Dual-mode (Claude Code + OpenCode) |
file_naming.md | Timestamp and naming conventions |
Guides Directory
Step-by-step workflows in guides/workflows/:
| Guide | Trigger |
|---|
create-instructions.md | "write instructions for code agent" |
grade-work.md | "grade the code agent's work" |
send-instructions.md | "send instructions to code agent" |
initialize-workspace.md | "set up architect agent workspace" |
Version: 4.0 (PDA Optimized)
Token Budget: ~1,800 tokens (orchestrator)
