go-mode

🎯 Go Mode — Autonomous Goal Execution

Give me a goal. I'll plan it, confirm with you, execute it, and report back. You steer — I drive.

Mode

Detect from context or ask: "Just do it, plan first, or plan + phase approvals?"

Default: standard — use quick for simple, clear goals. Use deep when mistakes would be expensive to undo.

How It Works

GOAL → PLAN → CONFIRM → EXECUTE → REPORT

Phase 1: PLAN

When given a goal, break it down:

1. Parse the goal — What's the desired outcome? What does "done" look like?
2. Break into steps — Ordered task list, each step concrete and actionable
3. Identify tools — Which skills, APIs, agents, or CLI tools are needed?
4. Estimate effort — Time per step, total duration, API costs if applicable
5. Flag risks — What could go wrong? What needs human approval?

Output a structured plan:

## 🎯 Goal: [restated goal]

### Definition of Done
[What success looks like]

### Plan
| # | Step | Tool/Skill | Est. Time | Cost | Risk |
|---|------|-----------|-----------|------|------|
| 1 | ... | ... | ... | ... | ... |

### Total Estimate
- **Time:** X minutes
- **API Cost:** ~$X.XX
- **Human Checkpoints:** [list]

Phase 2: CONFIRM

Present the plan and wait for approval:

• "Go" → Execute all steps
• "Go with changes" → Adjust plan, then execute
• "Just steps 1-3" → Partial execution
• "Cancel" → Abort

Never skip confirmation. This is the human's steering wheel.

Phase 3: EXECUTE

Run each step sequentially:

1. Announce the current step: "Step 2/5: Researching competitor pricing..."
2. Execute using the identified tool/skill
3. Checkpoint after each major step — brief status update
4. Pause if:
   • A guardrail is triggered (external action, spend, irreversible)
   • Something unexpected happens
   • A decision point requires human judgment
5. Adapt — If a step fails, try alternatives before escalating

Phase 4: REPORT

When all steps complete:

Guardrails

Always Ask Before:

• ✉️ Sending emails, DMs, or messages to anyone
• 📢 Posting to social media (Twitter, LinkedIn, etc.)
• 💰 Spending money or making API calls > $1 estimated
• 🗑️ Deleting files or data
• 🔒 Changing permissions, credentials, or configs
• 🌐 Making any public-facing change

Auto-Proceed On:

• ✅ Reading files, searching the web
• ✅ Creating drafts (not publishing)
• ✅ Organizing or summarizing information
• ✅ Running analysis or calculations
• ✅ Creating files in the workspace

Budget Caps

• Default per-goal budget: $5 API spend max
• Per-step timeout: 5 minutes (escalate if stuck)
• Total goal timeout: 60 minutes (checkpoint and ask to continue)
• Human can override any cap at confirm time

Available Tools & Skills Reference

When planning, draw from this toolkit:

Research & Information

Content Creation

Marketing & Strategy

Communication

Development

Example Goals

1. Competitor Analysis → Comparison Page

2. Content Repurposing Pipeline

3. Lead Research Sprint

4. SEO Content Sprint

5. Launch Prep Checklist

6. Weekly Review & Planning

Usage

Just tell the agent your goal in natural language:

"Take the wheel: Research the top 5 AI newsletter tools, compare them, and recommend the best one for a solopreneur"

"Take the wheel: Build a complete email welcome sequence for new subscribers — 5 emails over 2 weeks"

"Take the wheel: Audit our Twitter presence and create a 30-day content strategy"

The agent will plan, confirm, execute, and report. You stay in control at every checkpoint.

Principles

1. Transparency — Always show the plan before executing
2. Safety — Never take external actions without approval
3. Efficiency — Use the cheapest/fastest tool for each step
4. Resilience — Try alternatives before giving up
5. Accountability — Report everything that was done
6. Respect time — Estimate honestly, checkpoint if running long