# skill.md — Equip a Coding Agent (starter template from antist.ai)

> Skill-Centric Workflow: before wiring anything, your agent fills this file in.
> During the project, the agent KEEPS UPDATING it — new pitfalls, changed plans, done/undone.
> Any future agent (or future you) reads this file first and continues seamlessly.
> Method & updates: https://antist.ai/methodology · contact: Antisubmissivist@gmail.com

---
name: <agent-setup-name>
description: <one line: what this agent is for and whose machine it lives on>
status: planning   # planning | wiring | daily-driver
last_updated: <!-- agent: keep current -->
---

## 1. Goal & success criteria
- This agent's job: <e.g. "build and maintain my websites", "run my video pipeline">
- Success = you hand it a task in one sentence and review a finished, verified result
- NOT goals: <e.g. autonomous purchases, sending messages without approval>

## 2. Wiring shape decision
- Chosen: <chat-only | ide-copilot | terminal-agent | +phone-channel | multi-agent>
  (map: https://antist.ai/scenarios/equip-agent)
- Why: <one honest sentence>
- Upgrade signal: <the concrete sign it's time for the next shape>

## 3. The wiring (fill in exactly what's connected)
| Sense | Module | Status | Notes |
|---|---|---|---|
| Hands (files+shell) | | | permission mode? |
| Brain | | | subscription or API? budget/mo? |
| Memory | skill.md in <vault path> | | THIS file |
| Eyes (web/browser) | | | |
| Ears (voice) | | | |
| Remote channel | | | who is allowed to message it? |
| Tool protocol (MCP) | | | servers connected: |

## 4. Permission boundaries (write these BEFORE the first task)
- Agent may without asking: <read repo, edit files, run build/test, ...>
- Agent must ask first: <deploy, delete, send anything to another human, spend money>
- Agent may never: <secrets in plaintext, financial actions, ...>

## 5. Red lines
- Every task starts by reading this file; every task ends by updating it.
- Verified means VERIFIED — a checked box implies the agent ran/saw the result.
- Partial failures are reported loudly, never smoothed over.
- <!-- agent: append the red lines this setup earns the hard way -->

## 6. Known pitfalls (append as you hit them, with dates)
- <!-- example: 2026-07-09 agent CI-built artifact 500'd in prod; lesson: CI success ≠ healthy, verify content after deploy -->

## 7. Task log (newest first)
| Date | Task | Outcome | Verified how |
|---|---|---|---|
| | | | |

## 8. Open questions / next wiring upgrades
- <!-- agent: keep honest; delete only when done or explicitly dropped -->