v0.6.0 release

LightAgent v0.6.0 Release Notes

We are excited to announce the release of LightAgent v0.6.0 — a major step toward composable, skill‑driven agent development. This release completely refactors the core architecture, introduces native Skill support, and includes a built‑in secure sandbox for script execution. The result is a more modular, extensible, and task‑oriented agent framework. LightAgent is no longer just about "chat" — it's about getting things done.

📦 Upgrade Guide

pip install --upgrade lightagent>=0.6.0

🆕 New Features

• Native Skill Support – Package reusable capabilities (knowledge, workflows, tools) as portable Skills. Skills follow a progressive disclosure design, keeping the core lightweight while enabling complex, deterministic tasks. Now you can build once, reuse anywhere.

🧠 What’s Next

With v0.6.0, LightAgent evolves from a lightweight agent framework into a skill‑ready agentic platform. Whether you’re building research assistants, automation pipelines, or multi‑agent simulations — Skills let you package domain expertise into reusable modules, while MCP connects you to the wider tool ecosystem.

Try it in five minutes – and start shipping agents that actually remember, reason, and skillfully act.

What's Changed

• Docs: add multi-agent failure map (ProblemMap No.13) and link from README (#23) by @onestardao in #24
• fix: handle nested event loop in tool dispatch (fixes #11) by @awanawana in #28
• fix: prevent KeyError in get_history() when called before run() by @zhouraym in #30
• Fix: Remove invalid JSON argument replacement in non-stream tool call by @zhouraym in #31
• fix: TypeError when initializing LightAgent with debug=True and no log_file by @zhouraym in #29

New Contributors

• @onestardao made their first contribution in #24
• @awanawana made their first contribution in #28

Full Changelog: v0.5.0...v0.6.0