Claude Skills by scardoso-lu

Ultra-compressed communication mode. Cuts token usage ~75% by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman. Also auto-triggers when token efficiency is requested.

Ingest local staged files (CSV, Parquet, JSON, Excel) into a Microsoft Fabric Lakehouse as Bronze Delta tables. Use when loading data from the target repo's data/sandbox/ into the Bronze layer. Handles sanitization, lineage envelope injection, and idempotent partition overwrite.

Build Gold layer analytical objects — dimensional models (Kimball star schema), KPI aggregates, wide tables, or TMDL semantic models. Use when creating fact tables, dimension tables, monthly aggregates, or Power BI semantic models from Silver data. Enforces referential integrity, metric standardization, and schema locking.

Develop Fabric notebooks using a local closed-loop cycle — author in .py locally, build to .Notebook format, deploy via REST API, execute, monitor, diagnose errors, fix, and redeploy. Use for iterative notebook development without portal interaction. Typically converges in 1–3 iterations. Cold-start time varies by capacity tier (≈3 min on F64, up to 8–12 min on F2/F4).

Operate and maintain a Fabric data platform — orchestrate pipeline DAGs, run VACUUM, update platform inventory, manage workspace items, and execute operational routines. Use for day-to-day platform operations, maintenance windows, GDPR purges, or environment setup.

Create, deploy, and test a Fabric Data Factory pipeline that chains all topic notebooks in order (download → bronze → dq_bronze → silver → dq_silver → gold → dq_gold). Use after all notebooks for a topic have been individually deployed and smoke-tested.

Transform Bronze data into Silver — clean, deduplicate, enforce schema, and MERGE into Delta tables. Prefer SQL, fall back to SQL+Python hybrid for complex logic, use pure Python as a last resort. DQ checks run in a separate dq_silver_<source>.py notebook using Great Expectations.

Validate pipeline output using Great Expectations — row counts, null PKs, duplicates, schema drift, referential integrity, and business metric sanity. DQ notebooks run after ingestion notebooks and are the sole place where data quality is enforced. Produces a structured PASS/FAIL report.

Execute git commit with conventional commit message analysis, intelligent staging, and message generation. Use when user asks to commit changes, create a git commit, or mentions "/commit". Supports: (1) Auto-detecting type and scope from changes, (2) Generating conventional commit messages from diff, (3) Interactive commit with optional type/scope/description overrides, (4) Intelligent file staging for logical grouping

Interview the user relentlessly about a plan or design until reaching shared understanding, resolving each branch of the decision tree. Use when user wants to stress-test a plan, get grilled on their design, or mentions "grill me".

Generate deterministic synthetic CSV files under data/sandbox/ using the data_mock_generate MCP tool. Use when no real source file exists for a new or demo topic and you need staged data to build the download/bronze/dq notebook pipeline against.

Generate high-quality Product Requirements Documents (PRDs) for software systems and AI-powered features. Includes executive summaries, user stories, technical specifications, and risk analysis.

Prefix shell commands with rtk to apply the token-optimizing proxy (60–90% savings on Git, pytest, ruff, Fabric CLI, file ops).

List, read, and interpret Microsoft Fabric Semantic Models (Power BI datasets). Use when an agent needs to understand available business metrics, DAX measures, table relationships, or column definitions before writing DAX queries, building reports, or validating Gold-layer outputs against business definitions.