How it works
Powerful Support Skills
Agentic Bookmarks ships built-in skill guides — workflow playbooks that tell an AI agent when, why, and how to use the bookmark tools. Instead of every agent re-inventing good bookmarking practice, the know-how comes from the product itself.
No install, always current
The guides are delivered through MCP as dynamic resources — nothing is written into your project and nothing goes stale. The running server serves the current guide, and agents are pointed at the right one automatically as they reach the relevant tools. There's no separate installation to manage.
Slash commands aren't wired up yet — that's on the roadmap. For now you trigger a skill with a natural-language prompt, like the examples below.
Four skills at a glance
- Map the codebaseZero to a navigable map in a single pass.
- Analyze coverageTurn the bookmark set into a health report.
- Bookmark a subsystemCapture the shape of a named system.
- Bookmark specific filesPrecise annotation, no exploration.
More skills may come. If you have ideas for what would help your workflow, send feedback to contact@supermegalab.com.
Map the codebase
bookmarks://skill/map-codebaseZero to a navigable map in a single pass.
Point an agent at an unfamiliar repo and it builds a coherent, organized map: it surveys the major systems, drops semantic bookmarks at each one's entry points and load-bearing code, and groups everything so the result reads like a guided tour rather than a pile of pins.
It works system by system, deliberately, so groups stay coherent and generated or vendored code stays out. It's the fastest way to onboard yourself — or an agent — into a codebase nobody has mapped yet.
Try a prompt like
- “Index this whole project with bookmarks so I can find my way around later.”
- “I'm new to this repo — survey the major systems and bookmark each one.”
Analyze coverage
bookmarks://skill/analyzeTurn the bookmark set into a health report.
Reads the bookmarks you already have and hands back insight instead of a raw list: where coverage is thin, which areas are densely annotated, which anchors have gone stale or broken, and what themes recur across labels and groups.
It turns a bookmark set into situational awareness — a quick read on the health and shape of what's been captured, and a natural thing to run before starting work in an area.
Try a prompt like
- “What's already bookmarked here, and is any of it out of date?”
- “Give me an overview of our bookmark coverage before I dive in.”
Bookmark a subsystem
bookmarks://skill/add-to-systemCapture the shape of a named system.
Hand it a system — a directory, a feature area, a package — and it explores before it annotates: entry points, exported interfaces, the key algorithm, the non-obvious decision someone will trip over.
Everything lands under a single named group, so a subsystem's shape becomes durable, shareable knowledge instead of tribal memory.
Try a prompt like
- “Bookmark the authentication system.”
- “I'm about to work in packages/server — set up landmarks for it first.”
Bookmark specific files
bookmarks://skill/add-to-filesPrecise annotation, no exploration.
The lightweight one: no exploration phase, just direct annotation of files you already have in hand. Good when you're reading a file for another reason and want to capture what matters — exported symbols, load-bearing functions, the surprising workaround — before you move on.
Quality over quantity: a few focused bookmarks beat a wall of noise.
Try a prompt like
- “Bookmark the important parts of src/payments/checkout.ts.”
- “Annotate these three files with bookmarks for the key logic.”