Back to directory

Codemark

A command-line, semantic code-bookmarking tool that anchors a bookmark to a named structure such as a function using tree-sitter instead of a file-and-line reference, so it still points at the right place after a refactor, with a skill for agents.

Visit website (opens in a new tab)danielcardonarojas.github.io
Codemark

Codemark bookmarks code by structure rather than by line number. A file-and-line reference breaks the moment a line is added above it; Codemark uses tree-sitter to remember what was marked, such as a function, a block or a type, and anchors the bookmark to that named structure so it still resolves after the surrounding code is refactored or reformatted.

Bookmarks are grouped into named collections that record a flow through a codebase, such as the hops a request takes from an HTTP router to the database, with a short note on each bookmark's role; a later session, by the same person or a different agent, can load the collection and get a guided tour, a bug hunt over a known flow, or a comparison between two related flows.

It ships a skill that installs into a coding agent so the agent can create and recall these structural bookmarks on its own, and works with Claude Code, GitHub Copilot, Gemini CLI, and any agent that loads .agents/skills; it also has a native terminal dashboard. It is written in Rust and stores its data in SQLite with the sqlite-vec extension for semantic search.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.