Skip to content

Agent Workflow

AI agents work best when given structured context. ACE provides that structure at every step.

During implementation, the agent has access to:

  • charter.md — What to build (requirements, stories, criteria)
  • plan.md — How to build it (architecture, tech stack, constraints)
  • tasks.md — What to do next (ordered, phased task list)
  • constitution.md — Rules that cannot be broken
  • data-model.md — Entity definitions and relationships
  • contracts/ — API charters

The agent does not guess. It reads the charter, follows the plan, and executes the tasks.

ACE works with any AI coding agent:

AgentSlash Command SupportAgent File
Claude Code/auro.* commands.claude/CLAUDE.md
GitHub CopilotChat mode.github/copilot-instructions.md
CursorComposer.cursor/rules
WindsurfCascade.windsurf/workflows/
Gemini CLIDirect.gemini/commands/

Auro auto-detects your agent and generates the right file format.

Traditional documentation takes hours. ACE takes about 15 minutes:

  • 5 min: /auro.charter — write the charter
  • 5 min: /auro.plan — generate the plan
  • 5 min: /auro.tasks — break into tasks

After 15 minutes, you have a complete charter, a technical plan, and an executable task list. The agent knows exactly what to build, how to build it, and in what order.

After each phase, review the output:

  1. Do the tests pass? If not, fix before proceeding.
  2. Does the output match the charter? If not, the code is wrong.
  3. Do constitutional articles hold? If not, revise.

The agent marks tasks [X] as they complete. Incomplete tasks remain [ ]. You can see progress at a glance.