# Using Openspec Superpowers

> Use when starting, planning, implementing, reviewing, or closing a non-trivial OpenSpec change that needs durable specifications and disciplined agent execution.

- Skill: `jfear/using-openspec-superpowers` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add jfear/using-openspec-superpowers`
- Raw SKILL.md: https://api.skillmd.com/api/skills/jfear/using-openspec-superpowers/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- License: MIT
- Author: jfear (https://skillmd.com/u/jfear)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/jfear/using-openspec-superpowers

---


# Using OpenSpec with Superpowers

## Principle

OpenSpec governs durable change intent and current behavior. Superpowers governs execution. Keep one reviewed plan and one verification record with the active change; keep drafts and raw agent material isolated.

## Preflight

1. Read a project adapter skill when present; it supplies project checks and overrides generic defaults.
2. Read the active change's artifacts and `openspec instructions` before writing or implementing.
3. Ensure the project has the `agentic-workflow` schema. For a new project, inspect its conventions and obtain owner approval before installing or selecting the schema.
4. Ensure `.agent-work/{research,reviews,plans,specs,prompts}` exists in the current worktree. It must be ignored and must not be shared between worktrees.

## Artifact boundary

| Durable and committed | Ephemeral and ignored |
| --- | --- |
| `openspec/specs/` current behavior | `.agent-work/research/` |
| Active change proposal, delta specs, design, tasks | `.agent-work/prompts/` |
| Active change `plan.md` and `verify.md` | Draft plans, raw reviews, logs |
| ADRs with consequential rationale | |

Do not create root-level agent plans or reviews outside an active change. Promote a reviewed implementation plan to the schema-managed `plan.md`; summarize acceptance evidence in `verify.md`.

## Lifecycle

```text
explore -> feature worktree -> proposal/specs/design/tasks -> plan
        -> implement/test/reconcile -> verify -> sync -> human approval
        -> merge -> archive -> cleanup
```

- Explore read-only when scope is uncertain.
- Create a feature worktree before creating a named change. Never implement on `main`.
- Review all planning artifacts, including `plan.md`, before implementation.
- Follow TDD, task-level review, and project verification during implementation.
- If discoveries alter intent, reconcile the active OpenSpec artifacts; do not silently make code the new specification.
- Before final review, validate against the change and sync delta specifications into `openspec/specs/`.
- After merge, archive from the main worktree so only shipped changes enter history.

## Decision rules

If a correction is clearly implied by accepted intent, update the active artifacts and record the rationale. Ask the owner before a consequential architectural choice, ambiguity in intended behavior, irreversible/destructive operation, security-sensitive action, merge, push, archive, or deleting non-ephemeral material.

Urgency never bypasses branch isolation, review of planning artifacts, testing, or the durable/ephemeral boundary.

## Closeout

Complete `verify.md` with scenario coverage, commands/results, review conclusions, residual risks, and an explicit `READY` archive verdict. Before offering archive, inspect the active change and stop if `verify.md` is missing or its verdict is not `READY`; OpenSpec itself does not enforce this archive precondition. Never commit `.agent-work/`. Preserve the worktree until the owner chooses integration; then archive and clean it up with the project's supported worktree tool.