Using Superpowers Skill
Loads the dojo framework at session start: confirms the core disciplines are active, surfaces relevant lessons, and verifies clean state before any work begins. Does NOT replace the underlying skills — it sequences their loading and the session opening ritual.
When to Use
- At the start of every new session in a dojo-enabled repo.
- When the user says "use superpowers", "activate the dojo", or "start the framework".
- When another skill references the mandatory workflow.
- After resuming from a checkpoint or context handoff.
- NOT mid-session for already-active skills — re-loading wastes turns.
Prerequisites
- The dojo is installed in the repo (
skills/, optional-skills/, tasks/, spec/).
- The
view, edit, grep, and powershell Copilot tools.
git available to inspect branch + working tree.
tasks/lessons.md exists (created empty if first session).
scripts/verify.sh (or scripts/run-checks.ps1) available for the session-start gate.
How to Run
1. Review `tasks/lessons.md` for relevant active lessons.
2. Open `tasks/todo.md` — is there work in progress?
3. Confirm `git status`: right branch, clean tree (or known WIP).
4. Acknowledge: "Dojo framework active. Core disciplines loaded."
5. Hand off to the trigger skill for the user's actual request.
Quick Reference
| Always-on (core) |
Why it loads at every session |
behavioral-foundation |
The five non-negotiable prime directives |
plan-before-code |
Forces planning before any multi-step work |
verify-before-done |
Evidence required for any "done" claim |
self-improvement |
Lessons captured every correction |
demand-elegance |
Hacky solutions get challenged |
autonomous-bug-fix |
Full bug cycle without hand-holding |
safety-guardrails |
Destructive commands get a preflight check |
subagent-strategy |
Delegation is a first-class option |
| Workflow chain (practical) |
Triggered when |
brainstorming → plan-before-code → executing-plans → requesting-code-review → finishing-a-development-branch |
A non-trivial change begins |
| Workflow scenario |
Workflow shape |
| New feature |
Full pipeline: brainstorm → finish |
| Bug fix (non-trivial) |
Skip brainstorm, start at plan |
| One-line fix |
Direct fix + verify-before-done |
| Refactoring |
Start at plan, emphasis on test baseline |
| Code review |
Use code-review + receiving-code-review |
Procedure
Skill Activation Map
Always active (Core Disciplines):
behavioral-foundation — The five non-negotiable prime directives (the charter)
plan-before-code — Plan multi-step work before touching code
verify-before-done — Prove your work with evidence
self-improvement — Capture lessons, review at session start
demand-elegance — Challenge hacky solutions
autonomous-bug-fix — Full bug cycle, zero hand-holding
safety-guardrails — Preflight check before destructive commands
subagent-strategy — Delegate research and analysis
Workflow skills (activate in sequence):
brainstorming → using-git-worktrees → executing-plans → requesting-code-review → finishing-a-development-branch
On-demand skills (load when needed):
code-review — Reviewing PRs or diffs
refactoring — Safe restructuring
test-writing — Writing meaningful tests
pr-workflow — Preparing merge-ready PRs
debugging — Systematic complex debugging
codebase-onboarding — Understanding unfamiliar repos
dispatching-parallel-agents — Concurrent sub-agent work
receiving-code-review — Processing review feedback
skill-creator / writing-skills — Creating new skills
Session Start Ritual
Every session begins with:
- Read
memory/INDEX.md — What knowledge exists in the vault?
- Query relevant context —
bash scripts/memory-query.sh --type pattern --recent 5
- Review
tasks/lessons.md — Any un-promoted lessons?
- Check
tasks/todo.md — Is there work in progress?
- Verify git status — Are we on the right branch? Clean state?
- Acknowledge active skills — "Dojo framework active. Memory vault loaded. All skills live."
Session End Ritual
Every non-trivial session ends with:
- Write session summary —
memory/sessions/YYYY-MM-DD-summary.md
- Promote patterns — Any lessons at 3+ occurrences? Move to
memory/patterns/
- Record decisions — Any architectural choices? Log in
memory/decisions/
- Capture preferences — User corrections about style/approach? Log in
memory/preferences/
- Rebuild the graph —
bash scripts/link-index.sh
- Update lessons metrics —
tasks/lessons.md
Enforcement Rules
- All skills are mandatory — No skipping steps in the workflow
- No code without a plan (unless one-liner fix)
- No merge without verification —
scripts/verify.sh or equivalent
- No "done" without proof — Tests, logs, diffs
- Every correction becomes a lesson —
tasks/lessons.md
- Every session ends with a summary —
memory/sessions/
- Patterns are promoted, not buried — 3+ occurrences →
memory/patterns/
- Main branch is read-only — Work on feature branches/worktrees
Examples
Session Start:
Agent: "Dojo framework active. Memory vault loaded (4 decisions, 7 patterns,
3 preferences). Reviewed lessons.md (3 active lessons, 1 relevant to today's
task). No work in progress. On main branch, clean state. Ready."
Skill Triggering:
User: "I want to add real-time notifications"
Agent: brainstorming activates → asks Socratic questions → design approved
→ using-git-worktrees activates → creates feature branch
→ plan-before-code activates → writes plan to tasks/todo.md
→ executing-plans activates → works through tasks
→ requesting-code-review activates → self-reviews
→ finishing-a-development-branch activates → presents merge options
Pitfalls
- "I'll skip brainstorming, I know what to build" — You don't. Ask questions.
- "I'll just commit to main this once" — You won't "just" anything. Use a branch.
- "Tests take too long, I'll add them later" — Later never comes. Test now.
- "The plan is in memory only" — Write it in
tasks/todo.md or it doesn't exist.
- "I already know everything" — Review
tasks/lessons.md. You forgot something.
Verification
1---2name: using-superpowers3description: Activates the dojo framework at the start of a session.4tier: core5category: activation6created_by: human7platforms: [windows, macos, linux]8tags: [activation, framework, session]9author: Andreas Wasita (@andreaswasita)10metadata:11 source: https://skillmd.com/skills/andreaswasita/using-superpowers12---1314# Using Superpowers Skill1516Loads the dojo framework at session start: confirms the core disciplines are active, surfaces relevant lessons, and verifies clean state before any work begins. Does NOT replace the underlying skills — it sequences their loading and the session opening ritual.1718## When to Use1920- At the start of every new session in a dojo-enabled repo.21- When the user says "use superpowers", "activate the dojo", or "start the framework".22- When another skill references the mandatory workflow.23- After resuming from a checkpoint or context handoff.24- NOT mid-session for already-active skills — re-loading wastes turns.2526## Prerequisites2728- The dojo is installed in the repo (`skills/`, `optional-skills/`, `tasks/`, `spec/`).29- The `view`, `edit`, `grep`, and `powershell` Copilot tools.30- `git` available to inspect branch + working tree.31- `tasks/lessons.md` exists (created empty if first session).32- `scripts/verify.sh` (or `scripts/run-checks.ps1`) available for the session-start gate.3334## How to Run3536```text371. Review `tasks/lessons.md` for relevant active lessons.382. Open `tasks/todo.md` — is there work in progress?393. Confirm `git status`: right branch, clean tree (or known WIP).404. Acknowledge: "Dojo framework active. Core disciplines loaded."415. Hand off to the trigger skill for the user's actual request.42```4344## Quick Reference4546| Always-on (core) | Why it loads at every session |47|---|---|48| `behavioral-foundation` | The five non-negotiable prime directives |49| `plan-before-code` | Forces planning before any multi-step work |50| `verify-before-done` | Evidence required for any "done" claim |51| `self-improvement` | Lessons captured every correction |52| `demand-elegance` | Hacky solutions get challenged |53| `autonomous-bug-fix` | Full bug cycle without hand-holding |54| `safety-guardrails` | Destructive commands get a preflight check |55| `subagent-strategy` | Delegation is a first-class option |5657| Workflow chain (practical) | Triggered when |58|---|---|59| `brainstorming` → `plan-before-code` → `executing-plans` → `requesting-code-review` → `finishing-a-development-branch` | A non-trivial change begins |6061| Workflow scenario | Workflow shape |62|---|---|63| New feature | Full pipeline: brainstorm → finish |64| Bug fix (non-trivial) | Skip brainstorm, start at plan |65| One-line fix | Direct fix + verify-before-done |66| Refactoring | Start at plan, emphasis on test baseline |67| Code review | Use code-review + receiving-code-review |6869## Procedure7071### Skill Activation Map7273**Always active (Core Disciplines):**74- `behavioral-foundation` — The five non-negotiable prime directives (the charter)75- `plan-before-code` — Plan multi-step work before touching code76- `verify-before-done` — Prove your work with evidence77- `self-improvement` — Capture lessons, review at session start78- `demand-elegance` — Challenge hacky solutions79- `autonomous-bug-fix` — Full bug cycle, zero hand-holding80- `safety-guardrails` — Preflight check before destructive commands81- `subagent-strategy` — Delegate research and analysis8283**Workflow skills (activate in sequence):**84- `brainstorming` → `using-git-worktrees` → `executing-plans` → `requesting-code-review` → `finishing-a-development-branch`8586**On-demand skills (load when needed):**87- `code-review` — Reviewing PRs or diffs88- `refactoring` — Safe restructuring89- `test-writing` — Writing meaningful tests90- `pr-workflow` — Preparing merge-ready PRs91- `debugging` — Systematic complex debugging92- `codebase-onboarding` — Understanding unfamiliar repos93- `dispatching-parallel-agents` — Concurrent sub-agent work94- `receiving-code-review` — Processing review feedback95- `skill-creator` / `writing-skills` — Creating new skills9697### Session Start Ritual9899Every session begins with:1001011. Read `memory/INDEX.md` — What knowledge exists in the vault?1022. Query relevant context — `bash scripts/memory-query.sh --type pattern --recent 5`1033. Review `tasks/lessons.md` — Any un-promoted lessons?1044. Check `tasks/todo.md` — Is there work in progress?1055. Verify git status — Are we on the right branch? Clean state?1066. Acknowledge active skills — "Dojo framework active. Memory vault loaded. All skills live."107108## Session End Ritual109110Every non-trivial session ends with:1111121. Write session summary — `memory/sessions/YYYY-MM-DD-summary.md`1132. Promote patterns — Any lessons at 3+ occurrences? Move to `memory/patterns/`1143. Record decisions — Any architectural choices? Log in `memory/decisions/`1154. Capture preferences — User corrections about style/approach? Log in `memory/preferences/`1165. Rebuild the graph — `bash scripts/link-index.sh`1176. Update lessons metrics — `tasks/lessons.md`118119### Enforcement Rules1201211. **All skills are mandatory** — No skipping steps in the workflow1222. **No code without a plan** (unless one-liner fix)1233. **No merge without verification** — `scripts/verify.sh` or equivalent1244. **No "done" without proof** — Tests, logs, diffs1255. **Every correction becomes a lesson** — `tasks/lessons.md`1266. **Every session ends with a summary** — `memory/sessions/`1277. **Patterns are promoted, not buried** — 3+ occurrences → `memory/patterns/`1288. **Main branch is read-only** — Work on feature branches/worktrees129130### Examples131132**Session Start:**133> Agent: "Dojo framework active. Memory vault loaded (4 decisions, 7 patterns,134> 3 preferences). Reviewed lessons.md (3 active lessons, 1 relevant to today's135> task). No work in progress. On main branch, clean state. Ready."136137**Skill Triggering:**138> User: "I want to add real-time notifications"139> Agent: *brainstorming activates* → asks Socratic questions → design approved140> → *using-git-worktrees activates* → creates feature branch141> → *plan-before-code activates* → writes plan to tasks/todo.md142> → *executing-plans activates* → works through tasks143> → *requesting-code-review activates* → self-reviews144> → *finishing-a-development-branch activates* → presents merge options145146## Pitfalls147148- **"I'll skip brainstorming, I know what to build"** — You don't. Ask questions.149- **"I'll just commit to main this once"** — You won't "just" anything. Use a branch.150- **"Tests take too long, I'll add them later"** — Later never comes. Test now.151- **"The plan is in memory only"** — Write it in `tasks/todo.md` or it doesn't exist.152- **"I already know everything"** — Review `tasks/lessons.md`. You forgot something.153154## Verification155156- [ ] `memory/INDEX.md` was read and active context surfaced.157- [ ] `tasks/lessons.md` reviewed and any relevant rules acknowledged.158- [ ] `tasks/todo.md` checked for work-in-progress before starting new work.159- [ ] `git status` confirms correct branch and clean (or known-WIP) tree.160- [ ] Acknowledgement message issued ("Dojo framework active…").161- [ ] All Core Discipline skills are loaded for the session.162- [ ] Session-end ritual scheduled (summary + pattern promotion + link rebuild).