Skill Debate

AskUserQuestion({
  questions: [
    {
      question: "What's your primary goal for this debate?",
      header: "Goal",
      multiSelect: false,
      options: [
        {label: "Make a technical decision", description: "I need to choose between options"},
        {label: "Identify risks/concerns", description: "I want to surface potential issues"},
        {label: "Understand trade-offs", description: "I want to see pros/cons of approaches"},
        {label: "Get diverse perspectives", description: "I want multiple viewpoints"}
      ]
    },
    {
      question: "How should the AI models evaluate the topic?",
      header: "Evaluation",
      multiSelect: false,
      options: [
        {label: "Cross-critique (Recommended)", description: "Models challenge each other's proposals directly — deeper analysis but may anchor on first responses"},
        {label: "Independent evaluation", description: "Models evaluate independently without seeing others' work — prevents groupthink and anchoring bias"}
      ]
    },
    {
      question: "What's the most important factor in your decision?",
      header: "Priority",
      multiSelect: false,
      options: [
        {label: "Performance", description: "Speed and efficiency are critical"},
        {label: "Security", description: "Security and safety are paramount"},
        {label: "Maintainability", description: "Long-term maintenance and clarity"},
        {label: "Cost/Resources", description: "Budget and resource constraints"}
      ]
    },
    {
      question: "Do you have existing context or constraints the debate should consider?",
      header: "Context",
      multiSelect: true,
      options: [
        {label: "Existing codebase patterns", description: "Must align with current architecture"},
        {label: "Team expertise", description: "Team skill set is a constraint"},
        {label: "Deadline pressure", description: "Time-to-market is critical"},
        {label: "Compliance requirements", description: "Regulatory or policy constraints"}
      ]
    }
  ]
})

# Extract question and flags
QUESTION="Should we use Redis or in-memory cache?"
ROUNDS=3
STYLE="thorough"

# Dynamic advisor selection — the fleet builder remains the admission authority.
CONSULTATIVE_LIB="${HOME}/.claude-octopus/plugin/scripts/lib/consultative-advisors.sh"
ADVISOR_SELECTOR="${HOME}/.claude-octopus/plugin/scripts/helpers/select-fleet-advisors.sh"
source "$CONSULTATIVE_LIB" || exit 1
HOST_CLAUDE_ALLOWED=false
HOST_ADVISOR_SUCCESS=0
if octo_consultative_host_allowed; then
  HOST_CLAUDE_ALLOWED=true
fi
REQUIRED_EXTERNAL_ADVISORS=$(octo_consultative_required_external_count)
if ! ADVISORS=$("$ADVISOR_SELECTOR" debate standard "$QUESTION"); then
  echo "No eligible external debate advisors are available." >&2
  exit 1
fi

# Create debate directory structure
DEBATE_BASE_DIR="${HOME}/.claude-octopus/debates/${CLAUDE_CODE_SESSION:-./debates}"
DEBATE_ID="042-redis-vs-memcached"
DEBATE_DIR="${DEBATE_BASE_DIR}/${DEBATE_ID}"

mkdir -p "${DEBATE_DIR}/rounds"

# Write context.md
cat > "${DEBATE_DIR}/context.md" <<EOF
# Debate: ${QUESTION}

**Debate ID**: ${DEBATE_ID}
**Rounds**: ${ROUNDS}
**Style**: ${STYLE}
**Advisors**: ${ADVISORS}
**Started**: $(date -u +"%Y-%m-%dT%H:%M:%SZ")

## Question
${QUESTION}

## Clarifying Context

**Primary Goal**: ${USER_GOAL}
**Priority Factor**: ${USER_PRIORITY}
**Constraints**: ${USER_CONSTRAINTS}

## Additional Context
[Any relevant context from user's message or files]
[If claude-mem is installed, search for past debates or decisions on this topic using its MCP tools]
EOF

# Initialize state.json
cat > "${DEBATE_DIR}/state.json" <<EOF
{
  "debate_id": "${DEBATE_ID}",
  "question": "${QUESTION}",
  "rounds_total": ${ROUNDS},
  "rounds_completed": 0,
  "advisors": [$(echo "$ADVISORS" | sed 's/,/", "/g' | sed 's/^/"/' | sed 's/$/"/')],
  "user_context": {
    "goal": "${USER_GOAL}",
    "priority": "${USER_PRIORITY}",
    "constraints": "${USER_CONSTRAINTS}"
  },
  "status": "active",
  "created_at": "$(date -u +"%Y-%m-%dT%H:%M:%SZ")"
}
EOF

ORCH="${HOME}/.claude-octopus/plugin/scripts/orchestrate.sh"
DEBATE_PROMPT="You are {{advisor}} participating in debate round 1.

DEBATE QUESTION: ${QUESTION}

${CONTEXT}

Write a concise, independent analysis (${MAX_WORDS} words). Address implementation tradeoffs, risks, and where other likely perspectives may be wrong."
if ! SUCCESSFUL_EXTERNAL_ADVISORS=$(octo_launch_advisors "$ORCH" "$ADVISORS" \
    "${DEBATE_DIR}/rounds" r001_ "$DEBATE_PROMPT" "$REQUIRED_EXTERNAL_ADVISORS"); then
  echo "The required external debate advisors did not complete successfully." >&2
  exit 1
fi

Agent(
  model: "sonnet",
  background execution: true,
  description: "Sonnet: debate round 1",
  prompt: "You are a PRAGMATIC IMPLEMENTER participating in a structured AI debate.
YOUR ROLE: You are the person who would actually have to BUILD this. You care about what ships, what works, and what you'll be debugging at 2am. Ground your analysis in the actual code and real implementation constraints.

DEBATE QUESTION: ${QUESTION}

${CONTEXT}

Write your analysis (${MAX_WORDS} words) to: ${DEBATE_DIR}/rounds/r001_sonnet.md

Cover: implementation feasibility, hidden gotchas, concrete effort estimates, and what the other approaches miss from a builder's perspective."
)

if ! octo_consultative_provider_count_is_sufficient \
    "$SUCCESSFUL_EXTERNAL_ADVISORS" "$HOST_ADVISOR_SUCCESS"; then
  echo "A debate requires two successful, allowlisted providers." >&2
  exit 1
fi

# Read what all advisors said
for response_file in "${DEBATE_DIR}"/rounds/r001_*.md; do
  printf '\n## %s\n' "$(basename "$response_file" .md)"
  cat "$response_file"
done

# Write your analysis as moderator
cat > "${DEBATE_DIR}/rounds/r001_claude.md" <<EOF
# current host model Analysis - Round 1

[Your independent analysis here, considering but not just summarizing the three advisor perspectives. Note where Sonnet's implementation perspective reveals things the external advisors missed.]
EOF

evaluate_response_quality() {
    local response_file="${1}"
    local advisor="${2}"

    word_count=$(wc -w < "$response_file")
    has_citations=$(grep -c '\[' "$response_file" || echo 0)
    has_code=$(grep -c '```' "$response_file" || echo 0)
    addresses_others=$(grep -ciE '(gemini|codex|agy|antigravity|claude|sonnet)' "$response_file" || echo 0)

    score=0
    (( word_count >= 50 && word_count <= 1000 )) && (( score += 25 ))
    (( has_citations > 0 )) && (( score += 25 ))
    (( has_code > 0 )) && (( score += 25 ))
    (( addresses_others > 0 )) && (( score += 25 ))

    echo "$score"
}

for response_file in "${DEBATE_DIR}"/rounds/r001_*.md; do
    advisor=$(basename "$response_file" .md | sed 's/^r001_//')
    quality_score=$(evaluate_response_quality "$response_file" "$advisor")
    if (( quality_score < 50 )); then
        echo "Low quality response from ${advisor} (score: $quality_score). Re-prompting..."
        # Re-prompt for more detail
    fi
done

cat > "${DEBATE_DIR}/synthesis.md" <<EOF
# Final Synthesis: ${QUESTION}

## Summary of Perspectives

### External Advisor Perspectives
[Key points from each advisor selected in ADVISORS: Codex, Antigravity, or other available providers]

### 🟠 Sonnet's Perspective
[Key points from Sonnet across all rounds — especially implementation feasibility and gotchas]

### 🐙 current host model Perspective
[Your key points across all rounds]

## Areas of Agreement
[Where all advisors converged]

## Areas of Disagreement
[Key points of contention]

## Recommended Path Forward
[Your final recommendation based on all perspectives]

## Next Steps
[Concrete action items for the user]
EOF

I've completed a ${ROUNDS}-round debate on "${QUESTION}".

[Include key findings from synthesis.md]

Full debate saved to: ${DEBATE_DIR}

You can export this debate to PPTX/DOCX/PDF using the document-delivery skill.

User: /octo:debate Should we use Redis or in-memory cache?

Claude:
1. Creates debate folder at ~/.claude-octopus/debates/${SESSION_ID}/042-redis-vs-memcached/
2. Writes context.md with question
3. Round 1:
   - Launches Sonnet via Agent(model: sonnet, background execution: true) when the provider allowlist permits it
   - Calls orchestrate.sh spawn for each runtime advisor selected by build-fleet.sh, such as codex and agy
   - Waits for Sonnet completion
   - Writes own analysis (Opus) considering all advisor perspectives
4. Writes synthesis.md with final recommendation from all participants
5. Presents results in chat

User: /octo:debate -r 3 -d adversarial Review our authentication implementation in src/auth.ts

Claude:
1. Reads src/auth.ts to understand context
2. Creates debate folder
3. Round 1 (Sonnet launched in background first, then selected external advisors in parallel):
   - 🟠 Sonnet: Implementation feasibility analysis of auth.ts
   - External advisors selected by build-fleet.sh, such as 🔴 Codex or 🧭 Antigravity depending on availability
   - 🐙 current host model: Your independent analysis considering all advisors
4. Round 2:
   - 🟠 Sonnet: Responds to other participants' points
   - External advisors challenge each other's points
   - 🐙 Claude: You challenge advisor points
5. Round 3:
   - All participants: Final positions
6. Synthesis with quality scores for each advisor
7. Present results with cost tracking

After debate completes:
"Would you like to export this debate to PPTX/DOCX/PDF? I can use the document-delivery skill to create a professional presentation."

Knowledge mode "deliberate" phase → Run /octo:debate to get multiple perspectives
→ Use synthesis for final decision