brainstorm
Multi-persona Round-Table mit domain-aware Personas (LLM wählt 4–6 themenspezifische Rollen); iterative Cross-Pollination bis TF-IDF-Konvergenz; Synthesizer…
Skill by swDomass
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/swdomass-brainstorm
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: ml
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/brainstorm
What's inside
- Skill name
brainstorm- Description
- Multi-persona Round-Table mit domain-aware Personas (LLM wählt 4–6 themenspezifische Rollen); iterative Cross-Pollination bis TF-IDF-Konvergenz; Synthesizer ranked Top-N mit Pro/Contra
- License
- MIT
- Instructions
- 38 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
## System Prompt Addition
Du führst eine Brainstorming-Round-Table-Session mit dynamisch generierten Personas durch.
**Phase 0 — Topic-Analyse + Persona-Generierung (1 LLM-Call):**
Analysiere das Brainstorming-Thema und wähle 4–6 DIVERSE, themenspezifische Personas (keine generischen "Pragmatiker A/B"). Jede Persona bekommt einen unique `key` (kebab-case), `name`, `role_description`, `perspective_focus` und einen `system_prompt` ≥ 100 Zeichen — alles in EINER YAML-Struktur. Beispiele: Daten-Analyst + Boutique-Verkäuferin + Mitbewerber + Braut-Kundin für ein Pricing-Brainstorming.
**Phase 0.5 — Provider-Allocation (deterministisch, kein LLM-Call):**
Default: alle Personas auf Primary-Provider. Mit `#cross-provider`-Tag: Round-Robin über `(claude, gemini, codex, openrouter)` — degradiert sauber auf primary-only wenn keine Cross-Provider verfügbar sind.
**Phase 1 — Initial Idea Generation (1 Call pro Persona):**
Jede Persona produziert bis zu 10 Ideen unabhängig aus ihrer spezifischen Perspektive. Output strukturiert in ```` ```ideas ```` -Block mit Nummerierung. Quantität vor Qualität.
**Phase 2 — Cross-Pollination (1 Call pro Persona, iterativ):**
Jede Persona sieht die Ideen der anderen + ihre eigenen und contributes neue Ideen in 4 Kategorien: Aufbau-/Synthese-/Challenge-/Gap-Ideen. Keine Wiederholung eigener vorheriger Ideen.
**K — Konvergenz-Check (deterministisch, kein LLM):**
Greedy Single-Pass-Clustering der Ideen via Jaccard-Cosine (Threshold 0.40). Stop wenn `new_clusters / total < 20 %` UND mindestens eine vorherige Runde existiert (Runde 1 ist NIE konvergiert). Hard-Cap: 5 Iterationen.
**Phase 3 — Synthese + Ranking (1 LLM-Call, Primary-Provider):**
Wähle Top-N (default 5) aus allen Clustern. Berücksichtige Umsetzbarkeit, Originalität, Cluster-Größe und Diversität. Output: Markdown mit `## Top-N Ideen`, je Idee `Ursprung / Kern-Idee / Pro / Contra / Nächster Schritt`.
**Output:**
- Final-Report: `docs/brainstorm-YYYYMMDD-HHMMSS.md`
- State + Per-Iteration-Files: `.brainstorm/{ts}/`
- JSONL-Trace: `.brainstorm/traces/<run_id>.jsonl`
**Tags:**
- `#cross-provider` — Persona-Allocator über alle Provider verteilen (opt-in)
- `#max_iterations:N` (1–10, default 5)
- `#top_n:N` (1–20, default 5)
- `#min_personas:N` / `#max_personas:N` (2–10, default 4 / 6)
**Hard Rules:**
- Alle Persona/Synthese-Calls sind read-only — kein Tool darf Dateien im CWD modifizieren ausser dem finalen Report.
- Persona-Failures (eine Persona crashed) sind non-fatal — Tool macht mit Rest weiter, markiert Failure im Trace.
- Empty-Topic (alle Tags entfernt → leerer String) wird VOR dem ersten LLM-Call abgefangen.From swDomass/AI_orchestrator, MIT.
More from swDomass/AI_orchestrator
- 0.573score
review-loop
swDomass
Iterative code review — fix P1+P2 until clean, report P3 as an offer
SkillBFreshness: B★ 5Claude Code +9 - 0.573score
dev-loop
swDomass
Research → Execute → Dual-Review Loop (Code Quality + Issue Resolution) bis beide Reviews bestanden
SkillBFreshness: B★ 5Claude Code +9 - 0.535score
test-loop
swDomass
Run tests, fix failures, re-run until all tests pass
SkillBFreshness: B★ 5Claude Code +9 - 0.461score
deep-security-audit
swDomass
Multi-agent deep security audit — 6 expert personas (pentester, architect, code auditor, supply chain, data privacy, forensics) + CISO synthesis + optional…
SkillDFreshness: D★ 5Claude Code +9 - 0.461score
security-audit
swDomass
Security audit + fix — scans for hardcoded secrets, injection vectors, path traversal, input validation gaps, then implements all fixes
SkillDFreshness: D★ 5Claude Code +9 - 0.458score
knowledge-transfer
swDomass
Cross-Domain Know-How-Transfer — findet kreative Anwendungen von Vault-Expertise in anderen Branchen
SkillDFreshness: D★ 5Claude Code +9 - 0.458score
critical-review
swDomass
Radical-honesty architectural review — questions code, methodology, design, and operational fitness
SkillDFreshness: D★ 5Claude Code +9 - 0.450score
research-qa
swDomass
Deep Research + Fragen-Katalog vor Implementierungsbeginn
SkillDFreshness: D★ 5Claude Code +9
Signals
- Stars
- 5
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- May 16, 2026 (5 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/swDomass/AI_orchestrator
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from swDomass
- 0.573score
review-loop
swDomass
Iterative code review — fix P1+P2 until clean, report P3 as an offer
SkillBFreshness: B★ 5Claude Code +9 - 0.573score
dev-loop
swDomass
Research → Execute → Dual-Review Loop (Code Quality + Issue Resolution) bis beide Reviews bestanden
SkillBFreshness: B★ 5Claude Code +9 - 0.535score
test-loop
swDomass
Run tests, fix failures, re-run until all tests pass
SkillBFreshness: B★ 5Claude Code +9 - 0.461score
deep-security-audit
swDomass
Multi-agent deep security audit — 6 expert personas (pentester, architect, code auditor, supply chain, data privacy, forensics) + CISO synthesis + optional…
SkillDFreshness: D★ 5Claude Code +9 - 0.461score
security-audit
swDomass
Security audit + fix — scans for hardcoded secrets, injection vectors, path traversal, input validation gaps, then implements all fixes
SkillDFreshness: D★ 5Claude Code +9 - 0.458score
knowledge-transfer
swDomass
Cross-Domain Know-How-Transfer — findet kreative Anwendungen von Vault-Expertise in anderen Branchen
SkillDFreshness: D★ 5Claude Code +9
Cite this page
brainstorm skill. Agent Mart. https://agentmart.dev/skills/swdomass-brainstorm (last updated 2026-10-24)