# Posts by tag 'ai agents' — Liran Tal

AI security researcher, open-source maintainer, and author. Explore Liran Tal’s blog, research papers, developer tools, books, and conference talks.

Canonical: https://lirantal.com/tag/ai-agents

[![](/_astro/practical-second-brain-llm-markdown-knowledge-graph.CEcxBkpJ_Z19xT21.png)

Aug 2, 2026 17 min read

### The Practical Second Brain: Building an LLM-Operated Markdown Knowledge Graph

A hands-on guide to the files, schemas, workflows, governance, and maintenance routines behind an AI-assisted second brain.](/blog/practical-second-brain-llm-markdown-knowledge-graph)

*   [ai](/tag/ai)
*   [ai agents](/tag/ai-agents)
*   [llm](/tag/llm)
*   [knowledge management](/tag/knowledge-management)

 [![](/_astro/building-second-brain-llms-ai-agents.B3w2y6mA_2fR1l0.png)

Aug 1, 2026 16 min read

### Building a Second Brain with LLMs and AI Agents

How a plain-Markdown knowledge graph becomes a durable context layer, operating system, and compounding productivity tool for AI-assisted work.](/blog/building-second-brain-llms-ai-agents)

*   [ai](/tag/ai)
*   [ai agents](/tag/ai-agents)
*   [llm](/tag/llm)
*   [productivity](/tag/productivity)

 [![](/_astro/gemini-cli-invisible-unicode-skill-injection.AW0ivLkB_Z2nFK4g.png)

Jul 17, 2026 12 min read

### Invisible Unicode Can Hijack Gemini CLI Agent Skills

I disclosed a Gemini CLI prompt injection where invisible Unicode tag characters in SKILL.md caused the agent to run a hidden command. Google closed the report as out of scope because the workspace was trusted and YOLO mode was enabled.](/blog/gemini-cli-invisible-unicode-skill-injection)

*   [security](/tag/security)
*   [ai agents](/tag/ai-agents)
*   [gemini cli](/tag/gemini-cli)
*   [prompt injection](/tag/prompt-injection)
*   [unicode](/tag/unicode)

 [![](/_astro/is-prompt-injection-unsolvable-problem-ai.CNHPYK7E_1xccHK.png)

May 9, 2026 16 min read

### Is Prompt Injection an Unsolvable Problem in AI?

Prompt injection is not a bug you patch but an architectural limit baked into how LLMs process language. This post traces the problem from Gödel and Hofstadter to modern jailbreaks, agent skill supply chains, and why defenders must shift from elimination to survivability.](/blog/is-prompt-injection-unsolvable-problem-ai)

*   [security](/tag/security)
*   [ai](/tag/ai)
*   [prompt injection](/tag/prompt-injection)
*   [llm](/tag/llm)
*   [ai agents](/tag/ai-agents)

 [![](/_astro/managing-ai-agents-and-skills-with-apm.Cv8OZfsE_27cY3B.png)

May 2, 2026 14 min read

### Managing AI Agents and Skills with APM

Microsoft's Agent Package Manager brings package.json-style manifests and lockfiles to Claude Skills, MCP servers, and other agent artifacts, with hands-on recipes and sharp edges we hit in practice (collections path, meta-packages, security).](/blog/managing-ai-agents-and-skills-with-apm)

*   [ai agents](/tag/ai-agents)
*   [skills](/tag/skills)
*   [mcp](/tag/mcp)
*   [supply chain](/tag/supply-chain)