| name | supergoal |
| description | Principal-grade plan-only workflow for non-trivial software/product work. Builds a verified execution package before handing off to /goal or an implementation agent. |
| version | 1.0.0 |
| license | MIT |
| metadata | {"hermes":{"tags":["planning","execution","roadmap","autonomous-work","verification","workshop"],"related_skills":["writing-plans","test-driven-development","requesting-code-review"]}} |
SuperGoal
Use this when a project is too large, risky, or ambiguous for a one-shot answer.
SuperGoal is plan-only. It creates a rigorous execution package; it does not pretend the work is done.
Outputs
Create a .supergoal/ package with:
THINKING.md — objective, constraints, assumptions, risks.ROADMAP.md — phases with acceptance criteria.STATE.md — current status, blockers, next action.PROTOCOL.md — how the implementation agent should work and verify.PHASE-*.md — bite-sized phase specs.HANDOFF.md — exact command/prompt for the implementation agent.
Workflow
- Clarify the real objective and non-goals.
- Inspect the repo/system if available.
- Identify source of truth, deployment surface, secrets boundary, and tests.
- Split into small phases that can each be verified.
- Add kill criteria and rollback/stop conditions.
- Hand off to
/goal, a coding agent, or a human only after the plan has concrete acceptance tests.
Done criteria
A SuperGoal is done when another agent can start from HANDOFF.md without guessing what to build, where to build it, or how to prove it works.
