Releases: maxvaega/skillkit
Releases · maxvaega/skillkit
v0.4.0
v0.4.0
What's Changed
advanced progressive disclosure + basic documentation
- Advanced Progressive Disclosure with LRU content caching
- Mtime-based automatic cache invalidation
- Argument normalization for maximum cache efficiency
- Thread-safe concurrent invocations with per-skill asyncio locks
- Cache management API (get_cache_stats(), clear_cache())
- Performance: <1ms cache hits vs 10-25ms first invocation
- Memory efficient: ~2.1KB per cached entry
- 80%+ cache hit rate achievable
- Created
docs/folder with basic documentation - fixed
base_dirissue #5
v0.3.0
v0.3.0 Script Execution
- ✅ Script Execution Support (Python, Shell, JavaScript, Ruby, Perl)
- ✅ Automatic script detection (recursive, up to 5 levels)
- ✅ Security controls (path validation, permission checks, timeout enforcement)
- ✅ Environment variable injection (SKILL_NAME, SKILL_BASE_DIR, SKILL_VERSION, SKILLKIT_VERSION)
- ✅ LangChain script tool integration (each script exposed as separate StructuredTool)
- ✅ Parameters normalization to lower-case
- ✅ Comprehensive error handling and audit logging
- ✅ Cross-platform support (Linux, macOS, Windows)
- ✅ Backward compatible with v0.1/v0.2 (except ToolRestrictionError removed)
v0.2.0 async and multi-folder discovery
- ✅ Full async/await support (
adiscover(),ainvoke_skill()) - ✅ Multi-source skill discovery (project, Anthropic config, plugins, custom paths)
- ✅ Plugin ecosystem with MCPB manifest support
- ✅ Priority-based conflict resolution
- ✅ Fully qualified skill names (
plugin:skill-name) - ✅ Nested directory structures (up to 5 levels)
- ✅ Secure file path resolution with traversal prevention
- ✅ LangChain async integration
- ✅ Updated examples (async_usage.py, multi_source.py, file_references.py)
- ✅ Backward compatible with v0.1