nyldn/claude-octopus v8.41.0

v8.41.0 — Feature Adoption: Hooks, Agents, Multi-LLM Debate Integration

on GitHub

What's New

Multi-LLM Debate Integration

• /octo:embrace: Debate gates at Define→Develop and Develop→Deliver transitions
• /octo:plan: "Multi-LLM debate the plan" option before execution
• /octo:review: Debate mode for architecture decisions and high-severity findings
• /octo:security: Red team debate, full adversarial cycle, and debate-on-critical modes
• flow-define: "Multi-LLM debate" option for technical approach decisions

New Hook Events (3)

• PreCompact — preserves workflow state before context compaction
• SessionEnd — finalizes metrics, persists preferences to auto-memory
• UserPromptSubmit — classifies task intent for skill routing

Native Agent Definitions (10)

• security-auditor, code-reviewer, backend-architect, tdd-orchestrator, debugger, performance-engineer, frontend-developer, docs-architect, cloud-architect, database-architect
• Factory AI droids auto-generated with octo- prefix

Command Consolidation

• 13 thin wrapper commands removed (49 → 36 commands)
• Matching skills now user-invocable for direct access

Auto-Memory

• SessionEnd persists preferences to project-specific auto-memory
• SessionStart restores preferences from auto-memory
• Uses CLAUDE_PROJECT_DIR for correct project targeting

Other

• Native HTTP telemetry hook with shell fallback
• Mixed model routing via CLAUDE_MODEL env var
• Spec mode plan view integration (CC v2.1.70+)
• 90-test suite for v8.41.0 features

Full Changelog: v8.40.0...v8.41.0