Agentic Data Kit

Power BI development with Power BI Desktop open — driven by your AI assistant.

Skills, scripts and an independent visual reviewer that let Claude Code, Copilot, OpenCode, Cursor and Antigravity edit your model and report while you watch them change in Desktop.

npx @luccapinto/agentic-data-kit@latest init

1 agent · 9 skills · 3 workflows · 5 targets

---

Contents

• 🧠 What it is
• 📈 The Power BI loop (Desktop open)
• 🚀 Install
• ✅ Prerequisites for the Power BI loop
• 🎯 What's inside
• 🧪 Validation scripts
• 🔄 Multi-tool sync
• 🤝 Contributing
• 🙏 Credits & references
• 📄 License

---

🧠 What it is

Generic LLMs don't know your standards, but they already write clean code. So this kit captures only what a model can't supply on its own: how Power BI projects are laid out (PBIP, TMDL, PBIR), how to drive Power BI Desktop while it is open, real quality gates (Best Practice Analyzer, PBIR validation), and your brand. Power BI is the flagship; the kit also ships dbt naming conventions, documentation templates and presentation builders (HTML decks, editable PPTX). Everything is authored once in .agent/ and compiled into each assistant's native format.

---

📈 The Power BI loop (Desktop open)

You keep Power BI Desktop open on the .pbip. The assistant changes the model live and the report on disk, then reloads Desktop and looks at the result.

flowchart TB
    subgraph MODEL["Model & measures"]
        direction LR
        M1["Modeling MCP<br/>live change"] --> M2["Save in<br/>Desktop"] --> M3["run_bpa.py"]
    end

    subgraph REPORT["Report visuals"]
        direction TB
        R1["Edit PBIR files"] --> R2["validate_pbir.py"]
        R2 --> R3["Check status<br/>save if unsaved"]
        R3 --> R4["Reload"]
        R4 --> R5["Screenshot<br/>all pages"]
        R5 --> R6{"Visual<br/>review"}
        R6 -- "FAIL" --> R1
        R6 -- "PASS" --> R7(["Done"])
    end

    MODEL ~~~ REPORT

    classDef step fill:#FFFFFF,stroke:#6B7280,color:#111827
    classDef live fill:#F2C811,stroke:#B38F00,color:#111827
    classDef gate fill:#DBEAFE,stroke:#2563EB,color:#111827
    classDef review fill:#EDE9FE,stroke:#7C3AED,color:#111827
    classDef done fill:#DCFCE7,stroke:#16A34A,color:#111827
    class R1,M2 step
    class M1,R4,R5 live
    class M3,R2,R3 gate
    class R6 review
    class R7 done

Loading

_{Yellow: acts on the open Desktop (Modeling MCP, powerbi-report-author preview) · Blue: validation gate · Purple: independent review by pbi-visual-reviewer (read-only, fresh context).}

Visual loop in order: edit PBIR → validate_pbir.py → confirm Desktop has no unsaved changes (powerbi-report-author preview <folder> --host desktop --status, details.hasUnsavedChanges must be false) → --reload → --screenshot <dir> --all-pages → pbi-visual-reviewer.

Three pitfalls

1. Reload discards unsaved Desktop changes. Check the status first; save in Desktop if needed.
2. Reload only covers the report. Model and measure changes go through the Modeling MCP, not through file reload.
3. An edited theme may not reload. Rename the theme file (and its reference) to force a pick-up.

Details: desktop-bridge.md and modeling-mcp.md.

Desktop-closed alternative. No Desktop Bridge, an older Desktop, or CI? The disk-only path stays supported: edit TMDL/PBIR on disk, validate with the scripts, then reopen the project in Desktop. It is documented in pbi-report-layer-pbir › Alternative: Desktop closed and pbi-semantic-layer-tmdl › Alternative: Desktop closed.

---

🚀 Install

npx @luccapinto/agentic-data-kit@latest init

Pick your assistant in the prompt, or skip it with --target:

npx @luccapinto/agentic-data-kit@latest init --target cursor

--target	Installs	Agent (pbi-visual-reviewer)
antigravity	.agent/ (source of truth) + scripts/sync_agents.py	agents/ — Antigravity has no subagents: run the review in a fresh chat
copilot	.github/ (agents · skills · prompts · copilot-instructions.md); your other .github files are left alone	agents/pbi-visual-reviewer.agent.md, tools mapped to Copilot names
claude	.claude/ (agents · skills · commands · workflows · CLAUDE.md)	native subagent
opencode	.opencode/ (agents · skills · commands) + root AGENTS.md	mode: subagent, edit/bash denied
cursor	.cursor/ (rules/*.mdc · commands · skills)	none — run the review in a fresh chat with only the screenshots, spec and PBIR

OpenAI Codex reads the root AGENTS.md. Instructions are written in English; the assistant replies in your language.

---

✅ Prerequisites for the Power BI loop

Need	Why
Power BI Desktop (Windows), with File > Options > Security > Desktop Bridge → "Enable external tool access…" turned on	The Bridge lets the CLI reload and screenshot the open report
Node 20+	Runs the installer, the report CLI and the Modeling MCP
npm i -g @microsoft/powerbi-report-authoring-cli	powerbi-report-author (validate, preview, reload, screenshot); also installs the lower-level powerbi-desktop bridge CLI
Power BI Modeling MCP registered in your MCP client (npx -y @microsoft/powerbi-modeling-mcp@latest --start), then "Connect to '<file>' in Power BI Desktop"	Live model and measure changes
(optional) Tabular Editor 2	Exact BPA results; without it the Python fallback runs
Python 3.9+	Validation scripts (standard library only)

Without these you still get the Desktop-closed path and the scripts' fallbacks.

---

🎯 What's inside

	Count	What it is
🤖 Agent	1	pbi-visual-reviewer — independent review of report screenshots
🧩 Skills	9	On-demand knowledge modules that self-activate by description
🔄 Workflows	3	Slash commands: /plan, /validate-pbi, /document-dashboard
🎯 Targets	5	Antigravity · Copilot · Claude Code · OpenCode · Cursor

Agent. An agent only exists if it imposes something the model can't impose on itself: restricted tools, an independent fresh-context review, a different model, or an automatic trigger. pbi-visual-reviewer has read-only tools (Read, Grep, Glob) and never sees the author's reasoning. It checks screenshots + spec + PBIR for overlap, clipped text, empty/error visuals, alignment, theme consistency and titles, and returns PASS/FAIL per page and per visual.

Skills

Skill	Use it for
pbi-semantic-layer-tmdl	Author Power BI semantic models (TMDL) and drive them live through the Modeling MCP
pbi-report-layer-pbir	Edit reports as code (pages, visuals, themes) and run the Desktop reload/screenshot loop
pbi-quality-rules	Real Best Practice Analyzer via Tabular Editor 2, with a Python fallback
dbt-modeling-conventions	dbt CTE structure, stg_/dim_/fct_/mart_ naming, keys and tests
documentation-templates	Runbooks, data dictionaries, metric definitions, ADRs, Power BI catalog
applying-visual-identity	Apply a company brand from a DESIGN.md
building-html-presentations	Decks, scrollable decks, interactive sites, consulting PDFs + chart cookbook
generating-pptx	Native, editable PowerPoint via python-pptx
creating-agents-and-skills	Decide whether to add an agent, skill or nothing — and keep every tool folder in sync

Workflows

Command	What it does
/plan	Writes docs/PLAN-<slug>.md for a data task — plan only, no code
/validate-pbi	Runs the BPA and the PBIR check on a PBIP project
/document-dashboard	Generates Markdown documentation for a PBIP project

---

🧪 Validation scripts

Validation lives in scripts inside the skills, not in agents. All three accept --format text|json; exit code 0 = no errors, 1 = errors found, 2 = usage or tool failure.

Script	What it does
.agent/skills/pbi-quality-rules/scripts/run_bpa.py <model>	Real BPA through the Tabular Editor 2 CLI; falls back to bpa_fallback.py if TE2 isn't found
.agent/skills/pbi-quality-rules/scripts/bpa_fallback.py <model>	The same 10 rules of bpa-rules.json re-implemented in Python over TMDL (less exact)
.agent/skills/pbi-report-layer-pbir/scripts/validate_pbir.py <path>	Runs powerbi-report-author validate when installed; otherwise a structural PBIR check

Python 3.9+, standard library only. Run the repo's own tests with npm test.

---

🔄 Multi-tool sync

.agent/ is the single source of truth. One script compiles it into every tool's native format:

python3 scripts/sync_agents.py          # regenerate .claude/ .github/ .opencode/ .cursor/ AGENTS.md
python3 scripts/sync_agents.py --check  # exit 1 if generated folders are stale (used by CI)

.agent/                ← edit here
├── agents/            # 1 independent reviewer
├── skills/            # 9 on-demand knowledge modules (+ scripts, references)
├── workflows/         # 3 slash commands
└── rules/             # always-on rules (rules.md)

.claude/  .github/  .opencode/  .cursor/  AGENTS.md   ← generated (do not edit)

The generated folders ship in the package because init installs one folder per tool.

---

🤝 Contributing

See CONTRIBUTING.md: author in .agent/ only, sync, test — and apply the agent-survival rule before adding anything. Release history is in CHANGELOG.md.

---

🙏 Credits & references

• Power BI Desktop Bridge
• Power BI Report skill
• Power BI Authoring MCP server
• microsoft/skills-for-fabric (MIT) — skill powerbi-report-cli
• DESIGN.md follows the Google Labs spec (Apache-2.0). See THIRD_PARTY_NOTICES.md.

---

📄 License

MIT © Lucca Pinto — repository: luccapinto/agentic-data-kit

_{Built for data teams who'd rather ship than re-explain the basics.}