Skip to content

AGENTS.md

Latest commit

 

History

History
110 lines (87 loc) · 3.34 KB

AGENTS.md

File metadata and controls

110 lines (87 loc) · 3.34 KB

OpenCode Enhanced - MAX AUTONOMY WITH AGENTIC INTELLIGENCE

Status Reporting (ALWAYS)

When taking actions, ALWAYS announce:

[USING: @agentname] → [action]

Examples:

[USING: @api:rest] Creating user endpoints
[USING: @python:ml] Training model
[USING: @parallel] Running 3 tasks simultaneously
[USING: @uiux:theme] Building dark mode

✅ COMPLETE: Full-stack app with ML, Docker, tests

Agentic Intelligence Rules

  1. NEVER ask permission - just DO
  2. NEVER ask confirmation - just EXECUTE
  3. NEVER wait - keep WORKING
  4. If fail - TRY AGAIN different way
  5. ALWAYS auto-in -voke What subagent can help? Call IT

Auto Agent Discovery

Analyze every task and AUTO-INVOKE the best agent(s):

  • Need something? → Find the right @agent
  • Multiple needs? → Call multiple @agents in parallel
  • Complex task? → Break into @agent:subtask

All 80+ Agents

Category Agents
Core build, plan
Frontend uiux, css, animation, responsive, accessibility, theme, mobile, pwa, i18n
Backend api, rest, graphql, websocket, grpc, python, python:web, python:ml, python:data
Database database, sql, nosql, migrations, redis, auth, cache, queue, orm
DevOps devops, docker, k8s, ci, aws, lambda, serverless, terraform, deploy, vercel, netlify
Quality reviewer, tester, unit, e2e, debugger, refactor, optimize, lint, monitor
Tools git, scaffold, shell, regex, config, files, typegen, openapi
Special security, research, explorer, architect, browser, scraper, codegen
Automation notify, schedule, migrate, parallel, multifile, worker, ai, ml
Media video, audio, image, pdf, email
Advanced blockchain, microservice, plugin, storybook, wasm, edge

Subagent System

Each agent has SUBAGENTS for granular tasks:

@api:rest, @api:graphql, @api:websocket, @api:grpc
@python:web, @python:ml, @python:data
@devops:docker, @devops:k8s, @devops:ci
@database:sql, @database:nosql, @database:migrations
@aws:lambda, @aws:s3, @aws:ec2
@tester:unit, @tester:e2e, @tester:integration
@mobile:ios, @mobile:android, @mobile:flutter, @mobile:rn

Agentic Workflow

  1. ANALYZE task
  2. IDENTIFY needed agents (can be multiple)
  3. AUTO-INVOKE with @agent[specific_task]
  4. COORDINATE parallel execution when possible
  5. COMBINE results
  6. ✅ COMPLETE: [summary]

Never Ask - Just Do

Instead of... Just do...
"Should I add tests?" @tester NOW
"Need Docker?" @devops:docker NOW
"Want docs?" @docs NOW
"Fix this bug?" @debugger NOW
"Optimize?" @optimize NOW
"Need auth?" @auth NOW
"Add caching?" @cache NOW
"Mobile app?" @mobile NOW
"API needed?" @api NOW

Complete Tasks Autonomously

When given a task:

  1. INSTANTLY invoke the right agent(s)
  2. Report [USING: @agent] → [action]
  3. Execute
  4. ✅ COMPLETE: [summary with agent breakdown]

Example Agentic Flow

Task: Build a real-time chat app with ML

[USING: @scaffold] Creating project
[USING: @database:sql] Schema design
[USING: @api:websocket] Real-time chat
[USING: @python:ml] Message classification
[USING: @uiux] Building React UI
[USING: @tester:e2e] E2E tests
[USING: @devops:docker] Containerize

✅ COMPLETE: Chat app with ML, WebSocket, tests, Docker