---
name: dashboard
description: One read-only browser dashboard over the flow's records, the reference library's cards and any other component installed beside it and bound in the project — every page the component's own command-line answer, rendered. Use when the owner wants to browse the records or the cards, when asked to start or stop the dashboard, and when writing a new component that should add its pages to it. Reads no record or card file itself and writes nothing.
license: Apache-2.0
compatibility: Python 3.11 or later, standard library only; a browser. Optional - the `markdown` package for rendered bodies (paragraphs otherwise). Renders only components that answer `--json pages`, today the `flow` and `library` skills' CLIs.
metadata:
  version: "0.1.0"
  script: scripts/dashboard.py
---

# The dashboard

`python3 scripts/dashboard.py`, run by its path from the project root (`${CLAUDE_PLUGIN_ROOT}/skills/dashboard/scripts/dashboard.py` under Claude Code), serves one site on localhost, port 8765 by default, with one section per mounted component: the flow's records with the ranked list, each record with its links and back-references, the directions file and the validator's verdict; the library's cards with a ranked search, each card with our notes, extracts and citing sites, the candidates and watches and the check.
It is a renderer and nothing else.
Every page is the answer of the component's command with `--json`, so the command line and the dashboard cannot disagree; no work is shown and no path into the library's store is served.
It re-asks on every request, so a transition run in another terminal shows on the next reload.
It answers only a request addressed to it by name: `--host 0.0.0.0` serves other devices under this machine's names and its addresses, and `--allow-host NAME` adds a name a proxy or a tunnel uses; a page of another site never loads one of its pages in the background.

## What it mounts

Components are found in this order, the first mount of a name winning: a `--component CMD…` on the command line; `dashboard.toml` at the project root, with `[[component]]` tables naming a `command` list; the skill directories beside this one whose `component.toml` names a `command` and a `bindings` file, mounted when the bindings file exists at the project root; the `flow` and `library` commands on PATH when their bindings file exists.
So a skill adds its pages to the one dashboard by being installed beside it — in the same skills directory, or as a sibling plugin of the same marketplace, each plugin read at its newest version — and bound in the project, and `dashboard --list` prints what would be mounted.

## Writing a component

A component is any command that answers `<command> --json pages` with a page manifest and every other page as `<command> --json <subcommand> …`; `references/protocol.md` is the manifest's shape — the mount, the navigation entries, the link patterns that turn an id in any text into a link across components, and the routes with their views: `markdown`, `lines`, `list`, `record`, `tables`, `check`.
A component ships a `component.toml` beside its `SKILL.md` naming its command relative to the skill directory and the project's bindings file.
`scripts/dashboard.py --selftest` mounts a stub component and renders every view.