agentleFS
Haweee47Haweee47/powerbi-autopilot/.claude/skills/new-report/SKILL.md

new-report

Skill2 starsChanged 44 days agoMITView on GitHub

Start a new Power BI report in this repo from a pre-built pilot. Always asks first — purpose (dashboard, measure table, metric-check matrix, deep dive, fulfillment ops), design theme (ten presets), language and page layout — showing a wireframe and palette for each choice, then copies the matching pilot spec, remaps only the missing fields, generates and validates the PBIP. Use whenever the user asks for a new report, dashboard or table, however casually they phrase it.

2 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions2 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 2

    1. ai-search-guru/getcito-worlds-first-open-source-aio-aeo-or-geo-tool★ 432same name
    2. ansvisor/ansvisor★ 241same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository2 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 2 are distinctive, and 8 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 8

    Very specific

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

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

    Read templates/catalog.json.

    line 25
  2. Distinctive:

    Ask in this order; it is the order of how much each one changes the result:

    line 29
  3. Very specific:

    A brand colour: If the user names one, build the preset before asking the theme question, then offer it as the first option: python …

    line 42
  4. Very specific:

    Read only what it prints: the new spec path, a one-screen model summary, and the list of pilot fields missing in the target model.

    line 52
  5. Very specific:

    Do not open the pilot's generated .Report, the theme JSON, or TMDL files.

    line 53
  6. Very specific:

    Follow the RULE notes: last-year and target measures stop at the last data date (a naive SAMEPERIODLASTYEAR or a full-year budget gives …

    line 60
  7. Very specific:

    If a missing column is only used inside one display measure (e.g.

    line 61
  8. Very specific:

    A second report on the same model: add --reuse-map <the filled model-map.json>.

    line 63
  9. Distinctive:

    Keep rail text short (brand name and subtitle about 25 Latin or 12 CJK characters).

    line 76
  10. Very specific:

    If Power BI Desktop is available, capture every page before calling it done: powershell -ExecutionPolicy Bypass -File …

    line 91

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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

New report from a pilot

The point is token economy: a finished pilot already encodes layout, theme, formatting and page flow. You only choose it and change what differs. Never rebuild a report from scratch when a pilot fits.

0. Ask before you build — this is not optional

A request in plain words is not a complete brief. "매출 대시보드 만들어줘" or "make me a sales report" says nothing about who reads it, where it is shown, or in what language. Those choices change the output more than anything you would infer, and they cost the user one click each.

So: do not run any generator until the user has chosen. No "I'll start with navy and we can change it later" — changing it later means regenerating and re-capturing everything. Ask once, build once.

The only exceptions: the request already names the option (then don't ask that one), or the user explicitly says to pick for them (then state every default you chose in your reply, so they can correct one).

1. Ask once — a single AskUserQuestion call, up to four questions

Read templates/catalog.json. Every option there carries a preview — a wireframe for purposes and layouts, the palette and card shape for themes. Put it in the option's preview field so the user sees what they are choosing. Use name[lang] for the label and when[lang] for the description, in the user's language.

Ask in this order; it is the order of how much each one changes the result:

#QuestionOptions
1Purpose5 in purposes. A question holds 4, so offer the best fit first plus the three closest, and name the fifth in the question text. Warehouse, logistics or fulfillment-centre requests → fulfillment (it brings its own model and sample data)
2Theme10 in themes, grouped classic · showcase · practical. Offer the recommended one first and the three closest by when; list the rest in the question text so the user can type one under Other
3Languagelanguages in the catalog — English (default), 한국어, 日本語, 简体中文
4Page layout2 in frames: left rail (default) or top bar. Suggest the top bar for wide tables and wall screens

Judge the recommendation from the request before you ask, and put it first. A board deck leans navy or broadsheet; a wall screen leans midnight or carbon; print and month-end lean ledger; accessibility, projectors and colour-vision deficiency lean universal or contrast.

A brand colour. If the user names one, build the preset before asking the theme question, then offer it as the first option: python tools/brand_theme.py --id <id> --accent "#RRGGBB" --base <closest preset> then python tools/build_themes.py.

2. Copy the pilot (zero reading)

python tools/new_report.py --purpose <id> --theme <id> --lang <code> [--frame top] --name <Name> [--model <TMDL folder>]

Read only what it prints: the new spec path, a one-screen model summary, and the list of pilot fields missing in the target model. Do not open the pilot's generated .Report, the theme JSON, or TMDL files.

The user's own model (--model). The pilot's DAX is written for the reference model, so the tool also lists every column and measure the pilot needs that the model lacks, including those used inside its DAX, and writes model-map.json next to the spec: empty values plus _hints with the reference definition, the English name, where it's used, and RULE notes. Fill it in:

  • columns: the user's 'Table.Column'. measures: DAX in the user's model; keep the format the skeleton suggests.
  • Follow the RULE notes: last-year and target measures stop at the last data date (a naive SAMEPERIODLASTYEAR or a full-year budget gives wrong YoY and attainment).
  • If a missing column is only used inside one display measure (e.g. Orders PY), map that measure name instead of the column.
  • The generator stops if a value is still empty. Worked example: examples/05-own-model/.
  • A second report on the same model: add --reuse-map <the filled model-map.json>. Known values are prefilled and hints are written only for what's still empty (example 05: the other three pilots needed 5, 1 and 8 new values). Read only those.

No model yet? A folder of CSVs, a workbook or a database comes first: python tools/new_model.py --csv <folder> (also --excel, --odbc). It prints every relationship and dimension it inferred. Then come back with --model.

3. Edit only the new spec

  • Replace each field the tool listed as missing with the closest field from the model summary.
  • Rewrite headline sentences (measures → Headline …) for the new subject. Keep the "item + number" pattern so no grammar depends on the value.
  • Change titles and subtitles. Keep one language unless the user wants several ({"en": …, "ko": …}).
  • Number units live in templates/_shared/measures.<lang>.json; don't hard-code units in titles.
  • A short list in a small card ("weakest stores"): keep "top": N next to "sort" so only the rows that fit are shown.
  • Keep rail text short (brand name and subtitle about 25 Latin or 12 CJK characters). The generator warns when a line may wrap.
  • Layout, colors, fonts and visual formatting: don't touch. If something truly needs a new layout, add it to design-system/layouts/layouts.json and run python tools/build_layouts.py.

4. Build and check

python tools/design_check.py examples/<Name>/report.spec.json
python tools/generate_pbir.py examples/<Name>/report.spec.json
powerbi-report-author validate examples/<Name>/<Name>.Report

design_check.py applies the rubric items a script can settle (a KPI without a comparison, a chart with no title, a bar with no order, a table asked to show more rows than fit) and stops before anything is generated. The generator then stops before writing if a field doesn't exist. Fix the spec, rerun.

If Power BI Desktop is available, capture every page before calling it done: powershell -ExecutionPolicy Bypass -File tools\render_check.ps1 -Dir examples\<Name> opens it in the Store Desktop, refreshes and saves each page to out/render/. Build with --local-data when the report uses the bundled sample model, or Desktop finds no data. Validation passing does not mean the screen is right — this repo has had reports that validated with 0 errors and rendered blank.

5. Report back

Say which pilot, theme, language and layout you used, what you changed in the spec, the validator result, and whether you looked at the rendered pages. If you chose any option instead of asking, say which and why, so it can be corrected.

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.