Repository navigation
Skills
Skills let LightAgent discover reusable task instructions from directories that contain SKILL.md files. A Skill is useful when a task needs detailed procedures, domain-specific rules, reference files, scripts, or assets.
Default discovery looks under skills/:
skills/
market-research/
SKILL.md
references/
scripts/
assets/
You can pass custom directories:
agent = LightAgent(
model="gpt-4.1",
api_key="your_api_key",
base_url="https://api.openai.com/v1",
skills_directories=["skills", "company_skills"],
)SKILL.md must start with YAML frontmatter:
---
name: market-research
description: Use this skill for market research, competitor scans, and source-backed summaries.
---
# Instructions
Follow this workflow when the user asks for market research...Required frontmatter fields:
namedescription
On startup, LightAgent discovers Skill metadata only. When Skills exist, the agent receives metadata and Skill tools:
list_skillsactivate_skillexecute_skill_scriptread_skill_reference
The model can call activate_skill to load the full SKILL.md instructions into context when a user request matches a Skill description.
Optional directories:
-
scripts/: executable helper scripts for deterministic work. -
references/: documents the Skill can read on demand. -
assets/: binary resources the Skill can use.
Keep scripts narrow and auditable. A Skill should not need broad filesystem or network side effects unless the use case requires them.
agent = LightAgent(
model="gpt-4.1",
api_key="your_api_key",
base_url="https://api.openai.com/v1",
auto_discover_skills=False,
)- Keep descriptions specific so activation is predictable.
- Put long procedures in the body of
SKILL.md, not in the description. - Put large references in
references/and load them only when needed. - Prefer scripts for deterministic file conversion, parsing, or validation.
- Review third-party Skills before running their scripts.
LightAgent Wiki - see the repository, releases, and issues.
- Installation and Your First Agent
- Tools and Integrations
- Memory, Sessions, and RAG
- Workflows and Multiple Agents
- Agent Backends and Codex
- Durable Execution, Jobs, and Subagents
- Safety, Observability, and Evaluation
- Usage, Cost Estimates, and Budgets
- Deployment, Migration, and Verification
- 安装与第一个 Agent
- 工具和集成
- 记忆、会话与 RAG
- 工作流与多 Agent
- 执行后端与 Codex
- 持久执行、Jobs 与 Subagents
- 安全、观测与评测
- 用量统计、费用估算与预算
- 部署、升级与验收
- Quick Start
- Core Concepts
- API Reference
- Examples Cookbook
- Migration Guide
- Tools
- Tool Generator
- Memory
- MCP
- Skills
- LightFlow
- Tree of Thought
- Self-Learning
- Multi-Agent
- Tracing and Debugging
- Langfuse Observability
- Model Providers
- browser-use Integration
- Testing and CI
- Deployment Guide
- Architecture
- Security
- Known Limitations
- FAQ
- FAQ 中文
- Roadmap
- Release Process
- Contributing