nyldn/claude-octopus v8.0.0

on GitHub

Added

Opus 4.6 & Claude Code 2.1.32 Integration - Major release leveraging latest Claude capabilities:

• Claude Opus 4.6 agent type (claude-opus) for premium synthesis and strategic analysis

  • New get_agent_command/get_agent_command_array entries for claude --print -m opus
  • Model pricing: $5.00/$25.00 per MTok (input/output)
  • OpenRouter routing updated to anthropic/claude-opus-4-6 for complexity level 3

• Claude Code v2.1.32 feature flags

  • SUPPORTS_AGENT_TEAMS - Detects Agent Teams availability
  • SUPPORTS_AUTO_MEMORY - Detects auto memory support
  • AGENT_TEAMS_ENABLED - Reads CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS env var
  • Provider status display shows Agent Teams indicator when available

• New agent personas in agents/config.yaml

  • strategy-analyst (claude-opus) - Strategic analysis and market research
  • research-synthesizer (claude-opus) - Research synthesis and literature review

• Role mapping updates

  • strategist role maps to claude-opus for premium synthesis tasks
  • synthesizer role upgraded from gemini-fast to claude for better quality

• Auto Memory guidance in CLAUDE.md for persisting user preferences across sessions

Changed

• Skill description compression - All 43 skill descriptions reduced to single-line (<120 chars) to fit within 2% context budget (~4,000 tokens)
• Cost banner dynamically shows "Opus 4.6" or "Sonnet 4.5" based on workflow agents
• metrics-tracker.sh updated with claude-opus-4-6 pricing ($5.00/MTok)
• model-config.md updated with Opus 4.6 as premium option, Opus 4.5 marked legacy
• Provider CLAUDE.md documents Opus 4.6 vs Sonnet 4.5 routing guidance
• Version bump: 7.25.1 → 8.0.0

[7.24.0] - 2026-02-03

Added

Enhanced Multi-AI Orchestration - Four major features for improved developer experience:

• Smart Router (/octo) - Single entry point with natural language intent detection

  • Analyzes keywords and context to route to optimal workflow
  • Confidence scoring (>80% auto-route, 70-80% confirm, <70% clarify)
  • Routes to: discover, develop, plan, validate, debate, embrace
  • Supports all 6 workflow types with intelligent fallbacks
  • Issue: #13

• Model Configuration - Runtime model selection for cost/performance optimization

  • 4-tier precedence: env vars > overrides > config > defaults
  • /octo:model-config command for easy management
  • OCTOPUS_CODEX_MODEL and OCTOPUS_GEMINI_MODEL environment variables
  • Persistent configuration in ~/.claude-octopus/config/providers.json
  • Per-project or per-session model customization
  • Issue: #16

• Validation Workflow - Comprehensive quality assurance with multi-AI debate

  • 5-step workflow: Scope Analysis → Multi-AI Debate → Quality Scoring → Issue Extraction → Report Generation
  • 4-dimensional scoring: Code Quality (25%), Security (35%), Best Practices (20%), Completeness (20%)
  • Pass threshold: 75/100
  • Automated issue categorization (Critical, High, Medium, Low)
  • Generates VALIDATION_REPORT.md and ISSUES.md
  • Interactive questions for validation priorities and triggers
  • Issue: #14

• Z-index Detection - Browser-based layer analysis for design system extraction

  • Step 4.5 added to /octo:extract workflow
  • Detects all elements with explicit z-index and positioning
  • Identifies stacking contexts and conflicts
  • Generates layer hierarchy table and stacking context tree
  • Recommendations for z-index management
  • Graceful degradation when browser MCP unavailable
  • Issue: #15

Changed

• Updated plugin count: 30 → 31 commands (added /octo, /octo:model-config)
• Updated skill count: 42 → 43 skills (added skill-validate.md)
• Enhanced extract workflow with optional z-index analysis
• Improved visual indicators for all multi-AI workflows

Fixed

• N/A (no bug fixes in this release)

Dependencies

• Browser MCP (optional): Required for z-index detection in extract workflow
• jq: Required for model configuration management

Test Coverage

• Phase 1 (Model Configuration): 10 tests
• Phase 2 (Smart Router): 20 tests
• Phase 3 (Validation Workflow): 26 tests
• Phase 4 (Z-index Detection): 27 tests
• Total: 83 tests passing

Breaking Changes

• None - Full backward compatibility with v7.23.0