From superpowers-plus
Orchestrates research-brainstorm-review-prioritize-document cycle for designing new skill families from scratch. Produces prioritized roster, architecture decisions, infrastructure map, and blockers.
npx claudepluginhub bordenet/superpowers-plus --plugin superpowers-plusThis skill uses the workspace's default tool permissions.
> **Purpose:** Guide a structured domain design process for a new superpowers skill family.
Mandates invoking relevant skills via tools before any response in coding sessions. Covers access, priorities, and adaptations for Claude Code, Copilot CLI, Gemini CLI.
Share bugs, ideas, or general feedback.
Purpose: Guide a structured domain design process for a new superpowers skill family. Origin: Codified from the Call Review Domain design (21-step, 10-phase methodology, March 2026). Output: Prioritized skill roster, architecture decision, infrastructure map, wiki design doc + tracking page.
Wrong skill? Writing individual skills →
skill-authoring. Feature design →debate. Brainstorming ideas →brainstorming.
Announce at start: "I'm using the domain-design skill to structure this domain design exercise."
This is a 10-phase workflow. Per todo-enforcement.always.md, persist all 10 phases to TODO.md with #plan-domain-{name} tags BEFORE starting Phase 1. Mirror to MCP as supplementary.
Orchestrates the full cycle of designing a new skill domain (3+ related skills). Composes brainstorming, adversarial-search, innovation, and wiki-orchestrator into a 10-phase process.
| # | Principle | Why |
|---|---|---|
| 1 | Research before brainstorming | Prevents designing skills around inaccessible data |
| 2 | Documentation lies — verify with real queries | Wiki said tables existed; SELECT was denied |
| 3 | Every brainstorm gets a harsh review | 3 rounds × 2 = 6 phases of diverge/converge |
| 4 | P0 means zero external dependencies | If it needs a DBA ticket, it's P1 at best |
| 5 | Walking skeleton first | Prove the full pattern before building more |
| 6 | Kill aggressively | 75% attrition is normal and healthy |
| 7 | File blocker tickets immediately | Don't wait until build phase to discover blocks |
Three rounds of diverge (brainstorm) → converge (harsh review), bookended by research and documentation.
| Phase | Type | What Happens | Gate |
|---|---|---|---|
| 1. SCOPE | Setup | Name domain, trace system landscape, create wiki tracking page | — |
| 2. RESEARCH | Discovery | Per system: read docs, explore schemas, run real queries (don't trust docs) | All systems researched |
| 3. SYNTHESIZE | Analysis | Infrastructure map: Connected / Permission Denied / Not Connected | — |
| 4. BRAINSTORM R1 | Diverge | 3-5 personas × 2+ skill ideas each → candidate roster (10-15) | — |
| 5. HARSH REVIEW R1 | Converge | Value filter: kill duplicates, merge narrow skills. 30-50% attrition. | — |
| 6. BRAINSTORM R2 | Diverge | Map survivors to tables/APIs. Classify: INTEGRATE / BUILD / BLOCKED | — |
| 7. HARSH REVIEW R2 | Converge | ⛔ HARD GATE: Run real queries for EVERY data source. File blocker tickets immediately. | All sources verified |
| 8. BRAINSTORM R3 | Diverge | Tier (P0-P3), choose architecture, design walking skeleton. Skills ≤250 lines. | — |
| 9. HARSH REVIEW R3 | Converge | YAGNI check, scope check, walking skeleton minimality | — |
| 10. DOCUMENT | Output | Wiki: design doc + tracking page + handoff brief via wiki-orchestrator | — |
P0 = zero external dependencies, verified accessible. P1+ = blocked on filed tickets.
Before reporting "design complete": wiki design doc published, tracking page updated, all blocker tickets filed with links in both docs.
| Failure | Fix |
|---|---|
| Design produces too many skills (scope creep) | Enforce P0-only delivery — defer P1+ to backlog |
| Missing infrastructure requirements (tools, APIs) | Infrastructure map is mandatory output — blockers stop design |
| Design never converges after harsh review | Cap at 3 review rounds — escalate to user for tiebreak |
# Example: design a new skill domain
node ~/.codex/superpowers-augment/superpowers-augment.js use-skill domain-design