Agent Workflow/Task Request Template

README.md

The accompanying file in this Gist is a template used to kick off workflows/task requests with a Copilot, Antigravity, or other agent of your choice.

Steps To Use

1. Duplicate the template file (e.g., [.agent/workflows/@@.md])
2. In the newly named template file, replace all @ with project specifics:
   • @@@ | Description (top), Task Summary (bottom)
   • @@ | Filename (bottom)
   • @ | Overview (top), Acceptance Criteria examples (near top)
3. Update the ### Recreate section to suit the task.
4. Remove references to any unneeded ## MCP Servers for the given task.
5. Update project path in ### Verification Steps: > cd ~/projects/my-project/app/
6. Copy/paste the contents of the ## Original Prompt for this Workflow section into your agent's chat
   • Double check your agent model
   • Submit

Template Prep

This template is currently being used in a Nuxt web app project. To prep the template:

• Update the environment specifics to suit your project's stack.
• Add your own environment- and project-specific guides and docs:
  • .agent/rules/00-global-rules.md
  • .agent/rules/01-app-architecture.md
  • .agent/rules/02-coding-standards.md

_workflow-template.md

description
Bug/Feature/Update: @@@

Overview

@...

• See ## Agent Workflow sections for more info:
  • ### Recreate
  • ### Symptoms
  • ### Expectations
  • ### Acceptance Criteria

Agent Rules/Prep

• Read:
  • .agent/rules/00-global-rules.md
  • .agent/rules/01-app-architecture.md
  • .agent/rules/02-coding-standards.md

Guardrails

• Do not attempt self-annealing more than 3 times.
• Do not touch generated build artifacts under dist.

MCP Servers

Testing

• Playwright MCP
  • Server is up and running
• Tests should be considered before writing any code.

Browser Debugging

• Chrome DevTools MCP [^1]
  • Server is up and running
• The agent is to use the MCP server to verify fixes, or to continue debugging in the event a fix does not work.

[^1] Chrome DevTools MCP Server Follow-up

• The whole purpose of setting up the Chrome MCP server is so that agents can use it to self-diagnose. If I have to paste console log errors every time, tasks will take days.
• If you cannot access the Chrome MCP server, then tell me what I need to do so that you can access it because so far as I know, I just need to make sure it's running, which it is.
• As referenced in the Rules/Prep, the app's primary endpoints are at:
  • http://localhost:3000
  • http://localhost:3000/login

Agent Workflow

Recreate

1. Prep for '@...'

• Open a new browser window.
• Visit the log in page:
  • http://localhost:3000/login
• Fill in the login fields
  • Use the following dev-only debug button element:
  • <button id="debugAuthFormLoginGood"
• Double check the fields have expected values; compare with first entry in [app/app/helpers/fakes/fakes.ts].
  • If not, clear both fields click the button again, and double check again.
• Submit the form to log in:
  • <button id="loginSubmitButtonId"

2. Observe

• You are redirected to the home page.
• The home page has a 'logout' button.

Symptoms

@

Expectations

• Workflow-specific:
  • @
• General:
  • If any part of this workflow is too high-level, inform the dev with a list of "Gaps To Add".
    • The agent should attempt to educate itself and filling gaps before raising them to the dev.

Acceptance Criteria

• Workflow-specific:
  • @Example: No async or Promise in signatures
  • @Example: No await on APIs
• General:
  • No type errors
  • No lint errors

Verification Steps

• % cd ~/projects/my-project/app/
• % npm run lint
• % npm run typecheck

Original Prompt for this Workflow

Task Summary: @@@

Please review the following workflow:

• [.agent/workflows/@@.md]