I am the Technical Staff of Myself.

I am an AI Agent researcher and engineer building practical harnesses for agents: the runtime, tool-use, memory, web access, skill execution, context packaging, safety, and evaluation layers that turn agents from demos into working systems.

My work sits between agent research, infrastructure engineering, and hands-on product building.

By agent harness, I mean the environment around an AI agent: how it searches, reads, remembers, uses tools, runs modular skills, audits itself, evaluates outputs, and carries context across sessions. I build these systems directly, then use them to study how agents behave in real workflows.

Recently, I have been building a practical SkillOps and agent harness stack. This includes modular skills for web access, social and content extraction, transcription, research monitoring, persistent memory, skill security scanning, and agent self-audit. I also study and contribute around agent runtimes such as OpenClaw, OpenCLI, AgentBox, and Hermes Agent.

On the research side, I map and operationalize AI knowledge. I built an AI knowledge graph from industry reports, work on AI talent and job monitoring tools, and maintain research workflows that help agents follow fast-moving AI systems, companies, papers, and people.

I also build private experimental games, using games as a way to explore interaction design, simulation, character behavior, and agent-driven experiences.

Email: luosongred@gmail.com

Stack: Python / TypeScript / JavaScript / Shell / HTML/CSS

• AI Agent Harnesses: tool use, memory, execution, context packaging, safety checks, and evaluation loops for practical agents.
• SkillOps: a framework for designing, testing, and operating modular skills in personal AI agents.
• OpenClaw Skills: installable skills for search, scraping, transcription, monitoring, memory, security, and agent operations.
• AI Knowledge Graph: turning AI industry reports into an interactive ecosystem map.
• Private AI Games: experimental game systems exploring agents, interaction loops, and simulation-driven experiences.

• AgentBox: portable packaging for agent settings, skills, memory, and session context.
• OpenClaw: contribution and research workspace for personal AI agent runtime design.
• OpenCLI: a universal CLI hub for websites, Electron apps, and local tools.
• Hermes Agent: agent runtime study and exploration around self-improving agent systems.
• web-access: browser automation, web search, and direct HTTP access for agents.

• skillops-paper: research paper and artifacts for designing, testing, and operating modular skills in personal AI agents.
• skill-design-guide: methodology and QA patterns for building production-grade agent skills.
• persistent-memory: long-term agent memory through daily logs, structured facts, and compact session memory.
• skill-security-guard: security scanner for agent skills, including prompt-injection and unsafe-code checks.
• agent-self-audit: health checks for agent memory, skills, cron jobs, config, and runtime state.
• lobster-guard: identity and safety guardrails for agent behavior in group-chat settings.

• all-net-search-read: cross-platform search and content extraction across WeChat, Xiaohongshu, X, YouTube, Reddit, Bilibili, and Weibo.
• x-twitter-scraper: public X/Twitter profile, timeline, and full-tweet extraction without login.
• wechat-reader: WeChat Official Account article reading and Markdown conversion.
• weibo-scraper: public Weibo content extraction without login.
• bilibili-video: Bilibili video download and local transcription.
• xiaoyuzhou-podcast: Xiaoyuzhou podcast fetching, local transcription, and structured summarization.
• book-hunter: ebook discovery across Z-Library and Anna's Archive metadata sources.

• ai-knowledge-graph: interactive AI ecosystem graph built from industry reports and research material.
• ai-info-radar: AI news and resource discovery across media, blogs, podcasts, books, and social platforms.
• ai-talent-graph: academic AI scholar profiling with OpenAlex, arXiv, Semantic Scholar, and ORCID.
• ai-talent-radar: recruitment-oriented AI talent search combining research, GitHub, and Chinese platforms.
• agent-job-monitor: daily Agent/LLM job monitoring across major AI companies.

• Private AI games: experimental game systems exploring agents, interaction loops, simulation, and character behavior.
• AI Personality Test: a lightweight AI-culture quiz and deployable static web experience.
• NanoBanana-PPT-Skills: AI-generated PPT visuals, transitions, and video presentation workflows.
• Youtube-clipper-skill: semantic video clipping, bilingual subtitles, and content repackaging workflows.

This account is where I research and build the harness around AI agents: how they search, read, remember, use tools, run skills, evaluate outputs, and operate safely across real workflows.