SuperClaude-Org/SuperClaude_Framework v4.3.0

SuperClaude v4.3.0 — Implementation Fixes, Security Hardening & Claude Code Alignment

on GitHub

SuperClaude v4.3.0

Highlights

• Security hardened: Removed shell=True code execution vector, added SHA-256 integrity verification for downloads
• Implementation gaps filled: ConfidenceChecker, intelligent_execute(), ReflexionPattern mindbase — all placeholders replaced with real code
• Agent installation: superclaude install now deploys 20 agents to ~/.claude/agents/
• 62 new tests: Comprehensive coverage for execution engine (136 total, all passing)
• Claude Code gap analysis: Full feature mapping identifying skills migration, hooks integration, and plan mode as high-priority alignment opportunities
• Documentation overhaul: Version consistency, accurate feature counts, new integration guide

Security

• Remove shell=True with user-controlled $SHELL from subprocess calls (#536)
• Add SHA-256 integrity verification for docker-compose/mcp-config downloads (#537)

Added

• Agent file installation to ~/.claude/agents/ via superclaude install (#531)
• 62 new tests for ParallelExecutor, ReflectionEngine, SelfCorrectionEngine
• Claude Code integration guide (docs/user-guide/claude-code-integration.md)
• Claude Code gap analysis in KNOWLEDGE.md

Fixed

• ConfidenceChecker: 4 placeholder methods → real implementations
• intelligent_execute(): error capture + variable shadowing bug
• MCP env var flag: --env → -e (#517)
• ReflexionPattern: mindbase HTTP integration with fallback
• .gitignore contradictions and duplicate entries
• version.py synced from 0.4.0 to 4.3.0
• Japanese triggers → English in pm-agent.md (#534)

Changed

• Version bumped to 4.3.0 across all 17 files
• Feature counts corrected: 30 commands, 20 agents, 7 modes, 8 MCP servers
• CLAUDE.md: complete project structure with all extension points
• PLANNING.md, TASK.md, KNOWLEDGE.md: updated architecture and priorities

Full Changelog

v4.2.0...v4.3.0