From superpowers-plus
Generates a single high-conviction, radically innovative answer to what most accretively improves a project. Skips incremental ideas and bug fixes.
How this skill is triggered — by the user, by Claude, or both
Slash command
/superpowers-plus:innovationThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
> **Core question:** What's the single smartest, most radically innovative, accretive, useful, and compelling addition I could make to this project right now?
Core question: What's the single smartest, most radically innovative, accretive, useful, and compelling addition I could make to this project right now?
That question IS this skill. Everything below exists to answer it well.
Wrong skill? Incremental feature ideas → brainstorming. Bug fixes → systematic-debugging. Cleanup → pre-commit-gate.
Innovation mode: I've read the codebase and open issues.
What's the single most impactful addition we could make right now?
You are not generating a list. You are not filling out a rubric. You are answering ONE question with genuine depth, conviction, and specificity. The answer must be:
Before you can propose what to add, you must deeply understand what's there.
codebase-retrieval aggressively. Spend real time here.perplexity-research to understand what competitors and adjacent domains are doing. What's the state of the art?Ask these. One per message.
| # | Question | Why |
|---|---|---|
| 1 | What's the single biggest pain point or missed opportunity you see right now? | Grounds your answer in their reality, not your assumptions |
| 2 | Who benefits most from this project, and what do they wish it did? | Forces user-centric thinking |
| 3 | What constraints should I respect? (team size, timeline, tech stack, politics) | Keeps "radical" from becoming "impossible" |
If the user already answered any of these in their prompt, confirm rather than re-ask.
Fallback for non-answers: Do not deadlock waiting for answers that aren't coming. One retry per question is the maximum; after that, move on.
| User Response | Action |
|---|---|
| Silence (no reply after retry) | Proceed with Phase 1 context only. State assumptions. |
| "Just answer" / "you decide" | Proceed immediately. State assumptions explicitly. |
| Partial answer | Accept what's given, note what's missing, proceed. |
| Off-topic or echo | Rephrase once. If still off-topic, proceed with Phase 1 context. |
| Contradictory answer | Surface the contradiction, ask to clarify once, then pick the interpretation most consistent with Phase 1 context. |
When proceeding without full Phase 2 input, prefix your answer with: "Based on Phase 1 research and the following assumptions: [list]. Confidence is lower than it would be with direct user input."
Now answer the core question. ONE answer. Not three. Not five. ONE.
Your answer must include:
See references/output-template.md for format.
After presenting your answer:
"Want me to: (a) draft an RFC for this, (b) prototype the first move right now, (c) challenge this idea with
think-twiceto stress-test it, (d) generate alternatives viabrainstorming, or (e) turn this into a phased execution plan viaplan-and-execute?"
| Failure | What Went Wrong | Fix |
|---|---|---|
| Generated a list instead of an answer | Avoided having conviction | Pick one. Justify why it beats the others. |
| Idea is technically impressive but useless | Skipped understanding user pain (Phase 1 step 2) | Go back, read issues and complaints |
| Idea is obvious / incremental | Didn't research the market (Phase 1 step 3) | Invoke perplexity-research, look at adjacent domains |
| Idea ignores constraints | Skipped Phase 2 question 3 | Ask about constraints before answering |
| Idea doesn't build on existing strengths | Didn't understand the codebase (Phase 1 step 1) | Read more code, understand the architecture |
| "First move" is actually a 3-month plan | Confused planning with prototyping | Describe what you'd build in ≤5 days |
npx claudepluginhub bordenet/superpowers-plus --plugin superpowers-plusGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.