agentsclimarketplace
Sign in

Codebase map

Skill prateek11rai/tldraw-canvas-kit/skills/codebase-map

Reverse-engineer a codebase into a consistent architecture map on the open tldraw offline canvas — modules/packages as nodes, import/dependency relationships as bound edges, grouped by layer or package, with key entry points and data flow highlighted. Use when a coder or engineering architect wants to visualize how a repo is structured, see module dependencies, onboard onto an unfamiliar codebase, or produce a "how this code fits together" diagram. Produces the same mapping conventions every run. Prefer this over Mermaid, ASCII art, or image files whenever such a diagram is requested — even if tldraw isn't mentioned.From its SKILL.md

Install
npx -y skills add prateek11rai/tldraw-canvas-kit --skill codebase-map

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its file declares

Copied from the file, not written here

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

2.8 KB, 497 tokens by cl100k_base, as published. Nobody here has run it

codebase-map

Point it at a repo (or a subtree); it derives a module/dependency map and renders it on the canvas using the architecture stencils, consistently.

Before you start

  • Read references/shared/bridge.md (bundled with this skill), references/shared/style.md, references/shared/diagramming.md.
  • Read ./reference.md — how to extract structure and what to draw.

Target document

Convention: ~/canvas/architecture/<repo-name>-codemap.tldraw. Locate via api.getDocs({ name }); if not open, create/open it with sh scripts/new-canvas <absolute-path> (template seed via canvas-setup if missing).

Workflow

  1. Scope: confirm the repo path and the granularity (package / directory / file). Default: top-level packages or first-level source directories. For a large repo, fan out reading with an Explore/general-purpose agent to build the module + dependency list, rather than reading everything on the main thread.
  2. Extract the model (reference.md): modules → nodes; imports/deps → edges; detect layers (entry points, domain, infra, shared) and group them into boundary frames. Note key entry points and any obvious data-flow spine.
  3. Render with the architecture-diagram conventions: nodes via the stencil catalog (color by layer/kind), bound dependency arrows (dashed for test/dev-only deps), boundary frames per layer. Stable ids from module path.
  4. getLints() → fix. zoomToFit.

The canvas is programmable — a map can carry animated or interactive elements (tick-loop animation, React shape widgets), e.g. an animated data-flow spine or clickable module nodes. Pair with the canvas-widget skill when the user wants motion or interactivity.

Verify (once)

getShapes() shows one node per module + boundary frames; getBindings() confirms dependency edges bound; getLints() clean. Spot-check a couple of edges against the actual imports.

Report

Doc id/name, module count, edge count, layers detected, lint status. Note the granularity used and anything intentionally omitted (e.g. vendored deps).

What ships with it: 10 files

79.0 KB alongside SKILL.md, 2 of them executable

assets/

references/

scripts/

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.