Skip to content

Latest commit

 

History

History
82 lines (55 loc) · 1.63 KB

File metadata and controls

82 lines (55 loc) · 1.63 KB

Workflow Guide

1. Initialize

/agent-do init

This creates AI_COORDINATION/, the file-backed source of truth for the project.

2. Set Rules

/agent-do constitution no edits outside assigned scope; security gates are mandatory

Rules in CONSTITUTION.md are treated as hard constraints.

3. Specify

/agent-do spec Build a professional settings UI with responsive layout
/agent-do clarify
/agent-do checklist ux

The spec describes user value and requirements. It should not choose implementation details unless they are user-facing constraints.

4. Plan

/agent-do plan TASK-001
/agent-do tasks
/agent-do analyze-artifacts

The plan must be decision-complete. Executors should not need to choose architecture, scope, validation strategy, security posture, or design direction.

5. Assign And Handoff

/agent-do assign TASK-001 Antigravity "src/ui/**"
/agent-do handoff TASK-001 Antigravity

Any capable tool can be assigned. The role comes from AGENTS.md, TASKS.md, and CLAIMS.md.

6. Review

/agent-do review TASK-001
/agent-do code-review TASK-001
/agent-do security-review TASK-001
/agent-do design-review TASK-001
/agent-do rules-check TASK-001

The reviewer must be independent from the executor for the same task.

7. Repair Or Archive

If a gate fails:

/agent-do repair TASK-001

If all gates pass:

/agent-do archive TASK-001

Token Budget Discipline

Use:

/agent-do compact

The skill should prefer STATUS.md, task context packs, paths, line references, and concise findings over pasted source or chat history.