Logseq is my choice for personal knowledge management. It is an offline-first, local-only outliner 1. When I’m working in the terminal with Gemini CLI or Claude Code, I often need to capture ideas, check my TODOs, or append a note to my daily journal.
I built a skill to bridge this gap.
What it does
The skill allows the agent to interact directly with the Logseq graph on my machine. It follows the Agent Skills open standard 2, similar to the Strunk and White skill I wrote for prose review.
-
- Capture: “Add this to my journal” or “Remember this.”
- Task Management: “Show my tasks” or “Add a TODO for tomorrow.”
- Retrieval: Reading specific pages or searching recent journal entries.
The agent understands Logseq’s specific outline syntax—using tabs for nesting and bullets for blocks—and the :preferred-workflow :now workflow markers (NOW, LATER, DONE).
How it works
The skill is a markdown file with clear instructions on graph structure and file naming conventions. It uses standard unix tools like grep and find to navigate the journals directory efficiently.
Because the agent has direct access to the filesystem, it doesn’t need an API. It reads and writes the .md files in ~/Documents/Notes/Logseq/ directly.
Getting it
The skill is available as a GitHub Gist. To use it:
- Create a directory in your project:
.gemini/skills/logseq/(or.claude/skills/logseq/). - Download the
SKILL.mdfile into that directory. - Activate it by asking the agent to “Read my journal.”
The agent will then follow the instructions in the skill file to manage your notes and tasks without needing a plugin or third-party service.
-
Agent Skills - An open standard for defining tools and capabilities for AI agents. ↩