Codemark - Generate a semantic code bookmarking system for humans and agents

duncode · September 9, 2026, 1:39am

Codemark - Generate a semantic code bookmarking system for humans and agents (Cargo/Rust).

CLI for Codemark
Usage: codemark [OPTIONS] <COMMAND>

codemark tui

Use It With Your Agent

codemark install-skill --agent agents --scope user

.agents
└── skills
    └── codemark
        ├── examples.md
        ├── queries
        │   ├── common.md
        │   ├── csharp.md
        │   ├── dart.md
        │   ├── go.md
        │   ├── java.md
        │   ├── python.md
        │   ├── rust.md
        │   ├── swift.md
        │   └── typescript.md
        ├── queries.md.bak
        ├── SKILL.md
        └── templates.md

/create-skill “Create a new skill that combines both ~/.agents/skills/java-code-review/SKILL.md and ~/.agents/skills/codemark/ to bookmark critical code with notes, leave comments for tasks/tickets, what you would want to know if starting over tomorrow that aligns with the java code review report.”

skills
    └── java-review-codemark
        ├── references
        │   ├── codemark-setup.md
        │   ├── common-mistakes.md
        │   ├── examples.md
        │   ├── java-queries.md
        │   ├── report-template.md
        │   └── review-criteria.md
        └── SKILL.md

create-skill = https://github.com/NeoLabHQ/context-engineering-kit.git

duncode · September 9, 2026, 7:55pm

Any other language via dynamic WASM grammar loading — no recompilation needed.

gh repo clone tree-sitter/tree-sitter-cpp

duncode · September 11, 2026, 12:24pm

grammar.wasm

Agent Integrate Codemark - Code Review + Codemark
Grove - Structural sight for coding agents

──> cargo binstall grove-cst-cli
──> cargo binstall grove-explore-bin

──> grove fetch
registry host: https://raw.githubusercontent.com/Entelligentsia/grove-registry/v1

  agda         1.3.3 ✓ 3 files, 2102 KB
  bash         0.25.1 ✓ 3 files, 1326 KB
  c            0.24.2 ✓ 4 files, 630 KB
  c_sharp      0.23.5 ✓ 4 files, 5225 KB
  cpp          0.23.4 ✓ 4 files, 3354 KB
  css          0.25.0 ✓ 3 files, 125 KB
  embedded_template 0.25.0 ✓ 3 files, 5 KB
  go           0.25.0 ✓ 4 files, 212 KB
  haskell      0.23.1 ✓ 3 files, 3716 KB
  html         0.23.2 ✓ 3 files, 17 KB
  java         0.23.5 · cached
  javascript   0.25.0 ✓ 4 files, 402 KB
  jsdoc        0.25.0 ✓ 3 files, 29 KB
  json         0.24.8 ✓ 3 files, 5 KB
  julia        0.25.0 ✓ 4 files, 2574 KB
  ocaml        0.25.0 ✓ 3 files, 2456 KB
  ocaml_interface 0.25.0 ✓ 3 files, 2046 KB
  php          0.24.2 ✓ 3 files, 1033 KB
  python       0.25.0 ✓ 4 files, 447 KB
  ql           0.23.1 ✓ 3 files, 304 KB
  regex        0.25.0 ✓ 3 files, 20 KB
  ruby         0.23.1 ✓ 4 files, 2056 KB
  rust         0.24.2 ✓ 4 files, 1091 KB
  scala        0.26.0 ✓ 4 files, 4835 KB
  tsx          0.23.2 ✓ 3 files, 1411 KB
  typescript   0.23.2 ✓ 3 files, 1380 KB
  verilog      1.0.3 ✓ 3 files, 17710 KB

26 fetched · cache: /home/user/.cache/grove/grammars

──> tree .cache/grove/grammars
.cache/grove/grammars
├── agda
│   ├── grammar.wasm
│   ├── manifest.json
│   └── tags.scm
├── bash
│   ├── grammar.wasm
│   ├── manifest.json
│   └── tags.scm
├── c
│   ├── grammar.wasm
│   ├── locals.scm
│   ├── manifest.json
│   └── tags.scm
├── cpp
│   ├── grammar.wasm
│   ├── locals.scm
│   ├── manifest.json
│   └── tags.scm
...

duncode · September 16, 2026, 11:40pm

OCR → Codemark

Scan codebases with the ocr CLI and review findings with the codemark CLI, then bookmark and classify results. Includes two skills (ocr-scan-codemark, ocr-review-codemark) and two agents (ocr-scan-codemark, ocr-review-codemark) plus shared codemark and open-code-review resources.

It bridges two external CLIs:

ocr (@alibaba-group/open-code-review) — an AI code-review CLI that reads Git diffs (ocr review) or full files (ocr scan) and emits structured, line-level findings.
codemark (DanielCardonaRojas/codemark) — a structural bookmarking CLI that anchors notes to tree-sitter AST nodes, so bookmarks survive renames and refactors (unlike file:line).

The extension’s job is the glue in between: run ocr scan/ocr review, then convert every finding into a codemark bookmark filed under a single collection (a “tour”).

Prerequisites

Both agents stop and report if any of these are missing:

ocr CLI reachable on PATH — install with npm i -g @alibaba-group/open-code-review.
codemark CLI reachable on PATH — install with cargo install --git https://github.com/DanielCardonaRojas/codemark codemark-cli.

Thanks Hope you enjoy.