Przejdź do treści
Biblioteka skilli
AI i agenci · Planowanie

Roundtable Strategies

/Roundtable Strategies

Podpowiada strategie prowadzenia dyskusji wielu agentów: w kolejce, metodę Disneya, debatę, konsensus i sześć kapeluszy.

Repozytorium
spec2ship ↗
Licencja
MIT
Ostatnia zmiana repozytorium
12.07.2026

Ocena i bezpieczeństwo

90/ 100Klasa: dobry
Format15 / 25
Frontmatter z nazwą i opisem, poprawna nazwa, opis do 1024 znaków, nazwa zgodna z folderem.
Odkrywalność25 / 25
Opis mówi, kiedy użyć skilla, ma sensowną długość i jest w 3. osobie.
Treść35 / 35
Co najmniej 150 słów i do 500 linii, nagłówki, listy kroków, przykłady albo kod.
Higiena15 / 15
Bez zaślepek typu TODO, bez ścieżek Windows, z licencją.
Ryzykobrak sygnałów w skanie
Pliki w paczce
6
Skrypty
0
Komentarze HTML
0
Adresy URL w SKILL.md
0
Przestarzałość

Brak sygnałów. Repozytorium zmienione 12.07.2026, bez odwołań do wycofanych modeli i API.

Licencja

MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.

Co robi i kiedy

Podpowiada strategie prowadzenia dyskusji wielu agentów: w kolejce, metodę Disneya, debatę, konsensus i sześć kapeluszy.

Wyzwalacze z opisu

creativeinnovationnew featurecompareevaluatechooseurgentfast
Oryginalny opis (EN)

This skill should be used when starting a roundtable session, facilitating technical discussions, or orchestrating multi-agent deliberations. Provides facilitation strategies including standard round-robin, Disney creative method, structured debate, consensus-driven, and Six Thinking Hats. Auto-detect: When topic mentions 'creative', 'innovation', 'new feature' → disney. When topic mentions 'vs', 'compare', 'evaluate', 'choose' → debate. When topic mentions 'urgent', 'fast', 'quick', 'asap' → consensus-driven. When topic mentions 'comprehensive', 'thorough', 'all angles', 'deep analysis' → six

Jak działa

  1. Purpose
  2. When to Use This Skill
  3. Strategy Auto-Detection
  4. Available Strategies
  5. Strategy resolution hierarchy
  6. Workflow-Specific Defaults
  7. Strategy Selection Guide
  8. Configuration Integration
  9. Strategy File Structure
  10. Reference Files
Sekcje z nagłówków SKILL.md, w oryginale · 200 linii, 1 036 słów
Podgląd SKILL.md
---
name: Roundtable Strategies
description: "This skill should be used when starting a roundtable session, facilitating technical discussions,
  or orchestrating multi-agent deliberations. Provides facilitation strategies including standard round-robin,
  Disney creative method, structured debate, consensus-driven, and Six Thinking Hats.
  Auto-detect: When topic mentions 'creative', 'innovation', 'new feature' → disney.
  When topic mentions 'vs', 'compare', 'evaluate', 'choose' → debate.
  When topic mentions 'urgent', 'fast', 'quick', 'asap' → consensus-driven.
  When topic mentions 'comprehensive', 'thorough', 'all angles', 'deep analysis' → six-hats."
version: 1.3.0
---

# Roundtable Strategies

## Purpose

This skill provides facilitation strategies for s2s Roundtable sessions. Each strategy defines how participants interact, what phases the discussion follows, and how consensus is reached.

## When to Use This Skill

- Starting a `/s2s:roundtable` session
- Running `/s2s:specs` requirements gathering
- Running `/s2s:design` architecture design
- Custom brainstorming or decision-making sessions

## Strategy Auto-Detection

When no strategy is specified, analyze the topic for keywords to recommend appropriate strategy:

| Keywords in Topic | Recommended Strategy | Reason |
|-------------------|---------------------|--------|
| creative, innovation, new, brainstorm, ideation | **disney** | Creative ideation benefits from Dreamer→Realist→Critic flow |
| vs, compare, evaluate, choose, decide, between | **debate** | Comparing options needs Pro/Con analysis |
| urgent, fast, quick, asap, time-sensitive | **consensus-driven** | Speed-focused, converging quickly |
| comprehensive, thorough, all angles, deep, complete | **six-hats** | Complete analysis from all perspectives |
| *(no match)* | **standard** | Balanced default for most discussions |

### Auto-Detection Example

```
Topic: "Creative approach for new authentication feature"
        ^^^^^^^^         ^^^
        creative         new

→ Recommended: disney strategy
```

## Available Strategies

| Strategy | Best For | Phases | Participation |
|----------|----------|--------|---------------|
| **standard** | General discussions, balanced input | Single phase, multiple rounds | parallel |
| **disney** | Creative problem-solving, innovation | 3 phases: Dreamer, Realist, Critic | sequential (phases) |
| **debate** | Evaluating controversial options | 2 sides: Pro vs Con | parallel (per side) |
| **consensus-driven** | Fast decision-making | Converging rounds | parallel |
| **six-hats** | Comprehensive analysis | 6 thinking modes | sequential |

---

## Strategy resolution hierarchy

> Added in v1.3.0 (TECH-002 Phase 4). Full worked examples + D3 hierarchy decision rationale: `${CLAUDE_PLUGIN_ROOT}/skills/roundtable-execution/references/strategy-resolution.md` + ADR-0011 Phase 4 addendum.

`commands/roundtable.md` Phase 1 resolves the active strategy in this fixed order. First source that yields a value wins; subsequent sources are NOT consulted.

```
CLI --strategy <value>                                 (highest priority)
  ├─ if set → use it
  └─ else → .s2s/config.yaml.roundtable.strategy.by_workflow_type[{workflow_type}]
            ├─ if set → use it
            └─ else → profiles/{workflow_type}.yaml.default_strategy
                      ├─ if set → use it
                      └─ else → error "no strategy resolvable"   (defensive; should not trigger in normal use)
```

**Forced override** (wins over all): `profiles/{workflow_type}.yaml.strategy_constraints.forced == true`. Only `brainstorm.yaml` sets `forced: true` (forces `disney` regardless of CLI/config).

**D3 source roles**:
- `.s2s/config.yaml`: **user canonical** (runtime authoritative, project-level overrides)
- `profiles/{workflow}.yaml`: **plugin fallback** (consulted only when config.yaml lacks the key)
- this `SKILL.md` (table below): **human-facing documentation only**, NOT consumed at runtime

---

## Workflow-Specific Defaults

> **Authoritative source**: profile YAMLs in `${CLAUDE_PLUGIN_ROOT}/skills/roundtable-execution/profiles/{workflow}.yaml` (fields `default_strategy`, `participants.default`, `strategy_constraints`). The table below is a human-readable summary; if it drifts, the YAML is correct.

Each workflow has a recommended default strategy and participant set:

| Workflow | Default Strategy | Default Participants | Rationale |
|----------|------------------|---------------------|-----------|
| **specs** | consensus-driven | PM, UX-Researcher, BA, QA | Requirements need broad agreement, not debate |
| **design** | debate | Architect, Security, TechLead, DevOps | Architecture trade-offs benefit from Pro/Con analysis |
| **brainstorm** | disney (forced) | Variable (--participants) | Creative exploration requires Dreamer→Realist→Critic |
| **roundtable** | standard | Architect, TechLead (configurable) | Generic discussion, single topic, no opinionated default |

### Artifact Types by Workflow

> **Authoritative source**: profile YAMLs `artifact_types[]` (fields `prefix`, `session_key`, `is_primary`). In-session artifacts are listed here. `ADR-*` (design) is a Phase 3 output artifact written to `.s2s/decisions/`, not an in-session entry in `session.yaml.artifacts`.

| Workflow | Primary Artifacts (in-session) | Secondary Artifacts (in-session) | Phase 3 output artifacts |
|----------|-------------------------------|---------------------------------|--------------------------|
| **specs** | REQ-*, BR-*, NFR-* | OQ-*, CONF-*, EX-* | (none beyond session summary) |
| **design** | ARCH-*, COMP-*, INT-* | OQ-*, CONF-* | ADR-* (written to `.s2s/decisions/`) |
| **brainstorm** | IDEA-*, RISK-*, MIT-* | OQ-*, CONF-* | (none beyond session summary; ideas promoted to `.s2s/ideas.md`) |

### Strategy-Workflow Compatibility

| Strategy | specs | design | brainstorm |
|----------|-------|--------|------------|
| **standard** | ✓ | ✓ | - |
| **consensus-driven** | ✓ (default) | ✓ | - |
| **debate** | ✓ | ✓ (default) | - |
| **disney** | - | - | ✓ (forced) |
| **six-hats** | ✓ | ✓ | - |

**Note**: Brainstorm workflow ALWAYS uses disney strategy regardless of --strategy flag.

## Strategy Selection Guide

### Use `standard` when:
- You need balanced perspectives from all participants
- The topic doesn't require specialized phases
- Time is limited and you want direct input

### Use `disney` when:
- Exploring creative solutions
- Building new features or products
- You need to separate ideation from criticism

### Use `debate` when:
- Evaluating trade-offs between options
- The decision is controversial or polarizing
- You want to stress-test a proposal

### Use `consensus-driven` when:
- Speed is important
- The group needs to converge quickly
- Previous rounds have narrowed options

### Use `six-hats` when:
- Comprehensive analysis is needed
- You want to avoid groupthink
- The topic has emotional and logical dimensions

## Configuration Integration

Strategies are loaded via `config.yaml`:

```yaml
roundtable:
  strategy: "disney"  # Loads skills/roundtable-strategies/disney.md

  # Optional overrides (warning if conflicts with strategy)
  overrides:
    participation: "sequential"
```

## Strategy File Structure

Each strategy file (`standard.md`, `disney.md`, etc.) defines:

```yaml
defaults:
  participation: "parallel"  # How participants interact
  phases:
    - name: "phase-name"
      prompt_suffix: "Additional instructions for this phase"
      participants: "all"  # or specific list
  consensus:
    policy: "weighted_majority"
    threshold: 0.6
validation:
  requires_sequential_phases: false
  min_participants: 2
```

## Reference Files

- **`references/standard.md`** - Standard round-robin details
- **`references/disney.md`** - Disney Creative Strategy details
- **`references/debate.md`** - Structured debate details
- **`references/consensus-driven.md`** - Fast convergence details
- **`references/six-hats.md`** - De Bono's Six Hats details

## Quick Reference

### Participation Modes

| Mode | Behavior | Mitigates |
|------|----------|-----------|
| `parallel` | All respond simultaneously | Sycophancy, echo chambers |
| `sequential` | One at a time, sees previous | Enables building on ideas |
| `targeted` | Facilitator picks who speaks | Focused expertise |

### Consensus Policies

| Policy | Description |
|--------|-------------|
| `simple_majority` | >50% agreement |
| `weighted_majority` | Confidence-weighted votes |
| `unanimous` | All must agree |

### Escalation

When consensus fails after `max_attempts`, escalate to user with:
- All positions clearly presented
- Trade-offs explained
- Facilitator recommendation (if configured)
Pliki w paczce (6)
  • SKILL.md 9 KB
  • references/consensus-driven.md 5 KB
  • references/debate.md 6 KB
  • references/disney.md 5 KB
  • references/six-hats.md 8 KB
  • references/standard.md 3 KB

Wersje

W bazie nie ma innych skilli o nazwie Roundtable Strategies.

Migawka bazy z 13.11.2026 · wersja paczki baf6cceb30d7 (skrót SHA-256 całej paczki)

Pobrania w ostatnich 30 dniach: 0 · ranking

Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.

Podobne skille

  • 80Roundtable Execution
    Prowadzi wieloagentowe dyskusje w rundach z moderatorem i uczestnikami, np. przy specyfikacjach, projektach i burzy mózgów.
  • 100agent-roundtable
    Prowadzi wieloagentowy audyt funkcji, diffa lub specyfikacji, w którym Codex, Gemini, OpenCode i Claude debatują przez ACP, aż zbiegną się w specyfikację i listę zadań.
  • 100roundtable
    Zwołuje wieloagentowy okrągły stół Claude, Antigravity i Codex, które debatują pytanie, a decyzję o wyniku podejmuje użytkownik.
  • 95skill-debate
    Prowadzi ustrukturyzowaną debatę między Claude a innymi modelami, by sprawdzić ważne decyzje.
  • 100topic-debate
    Zestawia kilka AI w anonimowej debacie nad tezą, a niezależny sędzia wydaje werdykt i uzasadnienie.
  • 86team-shinchan:debate
    Uruchamia debatę wielu agentów, by znaleźć optymalne rozwiązanie, gdy trzeba porównać kilka propozycji.

Uwagi i zgłoszenia

Wiadomość trafi do Bartosza Sokolińskiego. Adres e-mail jest potrzebny tylko do odpowiedzi. Szczegóły w polityce prywatności.