菜单
Veteran-level reference for Power BI — PBIP project structure + git, semantic model design, DAX patterns + performance, deployment pipelines, refresh / gateway troubleshooting, integration with Power Platform solutions / ALM. Used by `power-bi-engineer` (primary).
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | power-bi |
| description | Veteran-level reference for Power BI — PBIP project structure + git, semantic model design, DAX patterns + performance, deployment pipelines, refresh / gateway troubleshooting, integration with Power Platform solutions / ALM. Used by `power-bi-engineer` (primary). |
Purpose: Provide veteran-level reference and patterns for the new power-bi-engineer agent — semantic model design, DAX, PBIP git workflows with Azure DevOps, deployment, refresh troubleshooting, and integration with Power Platform solutions/ALM.
SKILL.md for architecture and decision guidance.resources/*.md for specifics (PBIP structure, git patterns, DAX, etc.).solution-alm-engineer for end-to-end pipelines that include Power BI.resources/pbip-structure-and-git.mdresources/dax-patterns-and-performance.mdresources/deployment-and-pipelines.mdresources/refresh-gateway-troubleshooting.mdPower BI Copilot output quality is a function of how well the semantic model is grounded for AI, not of the report visuals — an under-grounded model yields thin, generic narratives because Copilot has to guess the goal and the preferred metric. When the work is "make Copilot produce meaningful reports," the lever is Microsoft's Prep data for AI feature — author in Microsoft's recommended order: AI data schema → verified answers → AI instructions → descriptions, then mark the model Approved for Copilot. The single highest-leverage goal channel is AI instructions (model-level, ≤ 10,000 chars): it states the model's purpose, maps user vocabulary → model fields, and defines business terms Copilot can't infer. The fill-in-the-blanks authoring artifact lives at ../../templates/power-bi-copilot-ai-instructions.md — drop a copy next to the model, replace the placeholders, paste each block into the surface it names. The field-level grounding floor (description/format/folder + synonyms) is ../../best-practices/enforce-measure-metadata.md.
For production-incident lessons (real customer engagements, with a decision tree + the workaround that resolved it), consult the plugin's knowledge/ directory in addition to the resources above. The full §8a table is in ../../CLAUDE.md; the Power BI–specific entry to know about:
knowledge/pbir-enhanced-report-loading.md — debug runbook. Read when a PBIR Enhanced report deploys but won't render. Covers the resourcePackages + version.json "infinite spinner" trap, the ~June 2026 Fabric breaking change rejecting prototypeQuery on visuals, and the correct nativeQueryRef + active projection shape. Verified against 7 real Enhanced repos 2026-06-02.knowledge/pbir-enhanced-reference.md — build reference. Read when authoring a new visual / page / report JSON. Full visual-type → queryState role catalog, filterConfig syntax (Categorical / Advanced / TopN / RelativeDate), slicer modes + pre-selection, objects vs visualContainerObjects split, literal-suffix rules, and verified real-file examples for KPI / Matrix / Scatter / Waterfall / Gauge / Multi-Row Card / 100% Stacked / Button Slicer. Synthesized 2026-06-02.knowledge/dax-category-name-mismatch-zero-scores.md — debug runbook + design pattern. Read when a scoring report deploys and renders fine but every measure returns 0 / BLANK across all entities — or proactively before writing any DAX measure that filters a dimension column by a hardcoded string. Captures the silent-zero failure shape, the SUMMARIZE diagnosis pattern, the Domain calculated-column fix (recommended) that decouples DAX from data-source category strings, and the score-scale gotcha. Production lesson from the BMA CSP Thematic Review (Contoso DEV, June 2026).knowledge/sempy-fabric-reference.md — Python-from-a-Fabric-notebook reference. Read when the work is semantic-model interrogation, DAX execution, refresh orchestration, TOM editing, or workspace / lakehouse / item enumeration via sempy.fabric. Inventory of every top-level function (workspace, dataset, refresh, TOM, lakehouse, item) + classes + submodules, with the ReadWrite-permission gate flagged. Distilled from the Microsoft Learn API reference 2026-06-02.This skill directly supports the new power-bi-engineer and ties into the enhanced ALM coverage for git/ADO scenarios.
Make VS Code terminal tabs show 🔔 + chime the moment an agent session needs you, across many parallel Copilot/Claude terminals. Installs three layers: workspace settings (tab bell icon + audio cue), a shell prompt hook (bell on command completion), and a background /proc-io watcher that rings a terminal's PTY bell when its agent process goes idle after responding. Use in a Codespace / VS Code setup where you run multiple background agent sessions and can't tell which one is waiting for input without clicking through each tab.
Validate a Dataverse create/update payload against LIVE entity metadata in ONE pass — before you POST/PATCH — so you never fix-one-field-and-retrigger. Catches nonexistent columns, invalid option-set values, malformed/missing lookup binds, missing required fields, and the owner-not-provided (SPN-create) trap together. Run it on the FIRST create failure, or up front when a human re-fire or long run gates each test.
Claude-Code-session-state surfacing contract for the Mímir dashboard tab. Documents the five reachable on-disk sources under `~/.claude/` and `<project>/.claude/`, the encoded-path algorithm + reverse-decode fallback, the hard scrubbing/torn-write/staleness/worktree/server-parity invariants, the per-card source map, and the honest empty-state contract for in-process-only fields. Read this before authoring or modifying `_read_mimir` (in both `serve-dashboards.py` copies), the `/__mimir` endpoint, the `#/mimir` generator tab, or the Gate 49 render fixture.
Author and validate the per-entity chart-of-accounts → statement-line mapping — the bespoke, judgment-laden asset that makes the close reusable per company and where mis-statements hide. Coverage-checked with statement_engine.py --lint-map. Used by `controller`.
Roll up N entity trial balances (same period) into a group consolidation — reuses statement_engine per entity, then applies a BALANCED intercompany-elimination journal so IC receivable/payable and IC revenue/COGS net to zero, emits an entity-columns + eliminations + consolidated worksheet, and flags (does not remeasure) non-functional-currency entities for CTA. Runs scripts/consolidate.py. 'Eliminate before you consolidate.' Used by `controller`.
Translate/remeasure ONE entity's functional-currency trial balance into a presentation-currency (USD) TB before consolidate.py, via scripts/remeasure.py. Current-rate method (A&L @ closing, P&L @ average, equity @ historical; plug = CTA to OCI/equity) or temporal method (monetary @ closing, non-monetary @ historical, P&L @ average except REV_EXP_HIST @ historical; plug = remeasurement G/L to net income). Reuses statement_engine's section-signed presentation + reasoning trail — no new sign logic. Blocks on a missing/invalid rate_class, blocks when the balancing plug fails the analytical CTA self-check, and refuses a highly_inflationary + current_rate combo (ASC 830 vs IAS 29). Decision-support, not an audited remeasurement. Used by `controller`.