agentleFS
AhmedQureshi89AhmedQureshi89/lean-experiment-kit/.claude/skills/sync-dashboard/SKILL.md

sync-dashboard

Skill0 starsChanged 19 days agoMITView on GitHub

Push the experiments folder to the founder's live dashboard (a Claude Dashboard artifact with four tabs - Today, Board, Model, Log), creating the dashboard on the first run. Use when the founder says "sync the dashboard", after /new-experiment, /log-experiment or /pull-evidence, or when the Stop hook says a sync is due.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Write that url into kit.config.json as dashboard_url straight away, so a failure below does not create a second dashboard next time.

    line 36
  2. Distinctive:

    If the page changed, run the check of 2.5.

    line 67
  3. Very specific:

    After the batch succeeded, run python dashboard/mark_synced.py.

    line 71

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

The dashboard is a Claude Dashboard artifact (Anthropic's built-in type), private to the founder until they share it. Its link is dashboard_url in kit.config.json. It reads four attached data files built from this repo (experiments, canvas, canvas-meta, events) and, for a running card that names one, a live query (live_dataset). Its page is dashboard/dash-index.html.

1. Build (every run)

In the repo root:

python dashboard/build_data.py
python dashboard/build_dash_files.py

The first reads every test card header and both canvases, rewrites experiments/PROGRESS.md (never hand-edit it) and writes the bundle. A PROBLEM: line means a header or canvas row is malformed: fix that file and run it again. Nothing is pushed until the build is clean. The second writes dashboard/out/experiments.json, canvas.json, canvas-meta.json, events.json and index.doc.json (the page, with kit.config.json's name, time zone and weekend written into it).

2. First run only: create the dashboard (dashboard_url is empty)

  1. Artifact publish with type_url: "https://claude.ai/artifact/GZLZthZ2sQjn7T16skfdeF" (the Dashboard type), title = "<business> experiments" from kit.config.json, auto_open: "after_first_write", and nothing else. The result carries the new dashboard's url and the type's own instructions: follow those for anything this skill does not say. Artifact types are set per account, so if that link is refused, run Artifact action: "quickstart" with intent: "other" and use the Dashboard type it names; if none is offered, tell the founder the account has no Dashboard type and stop.
  2. Write that url into kit.config.json as dashboard_url straight away, so a failure below does not create a second dashboard next time.
  3. Upload the four data files: Artifact with that url, asset: true, file_path = each JSON file in dashboard/out/ except index.doc.json, one call per file. Keep each reply's /_blob/… url.
  4. Load ArtifactData (ToolSearch query select:ArtifactData) and write one batch on the url:
    • set dash/meta {title: "<business> experiments"}
    • for each of the four files, set datasets/<id> (experiments, canvas, canvas-meta, events) with {title, description, source: {kind: "file", url: "<its /_blob url>", name: "<file name>"}, updated: {at: "<now, ISO with the config's utc_offset>", by: "Claude, from the experiments repo"}, schedule: "Claude re-syncs it whenever a card changes", staleAfterHours: 168}. Titles: "Experiment board", "Canvas boxes", "Canvas version and riskiest box", "What changed, by date". Descriptions: one sentence each on what the file holds, ending "Synced by Claude from the repo."
    • set files/index.html with file_path = dashboard/out/index.doc.json (it holds {text}).
  5. Check: ArtifactData get collection data/users/me, doc check. While it says checking, read again (5 reads at most). If it says failed, fix what it names in dashboard/dash-index.html (in the repo, so the fix is kept), build, write the page again, and check again (2 fixes at most).
  6. Tell the founder the link, and that only they can see it until they share it.

3. Every later run: point the datasets at new copies

  1. Upload the four data files as in 2.3.
  2. ArtifactData list collection datasets on the url; note each dataset's version.
  3. One batch: for each of the four, update datasets/<id> with if_version and data: {source: {kind: "file", url: "<new /_blob url>", name: "<file name>"}, updated: {at: "<now>", by: "Claude, from the experiments repo"}}. If dashboard/dash-index.html changed since the last sync, add set files/index.html with file_path = dashboard/out/index.doc.json and the if_version from a get of it. On a version conflict, read the versions again and retry once; if it conflicts again, stop and say so.
  4. If the page changed, run the check of 2.5.

4. Record and report (every run)

After the batch succeeded, run python dashboard/mark_synced.py. That stamp is what the Stop hook (dashboard/check_sync.py) checks, so a session cannot end with the dashboard behind the cards without being told. Then tell the founder in one line: synced, N experiments, the canvas versions, the link.

Each upload and each batch asks the founder to approve it. A session with nobody to approve (a scheduled or headless run) cannot sync: say so, and leave it to the founder's next /sync-dashboard. Old uploads are not deleted; the dashboard's Sources panel shows storage and can prune them.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.