Prompts & Resources
The posts show slices. This is the library: the complete instruction pages, agent briefs, validation queries, and notebook patterns, ready to copy. Every artifact links back to the post that explains why it's shaped the way it is.
Notion Agent Instructions
The instruction pages my Notion assistants run on: purpose, division of labor, hand-off rules, and hard stops.
-
Polpette: Consultant's Consultant Agent Instructions
markdown from Meet My Assistants: The Three Agents That Keep My Projects Honest
The full instruction page for my analysis and advisory agent, including the routing rules and the meeting review state machine.
Open →
Claude & MCP Instruction Pages
Operational briefs for execution sessions. Role, dated priority callout, standing directives, verified state, report-back requirement.
-
Fabric Backend Builder Brief
markdown from Fabric Is Best as a Backend
The brief an agent runs when the job is standing up an app backend in Fabric: code as the output, workspace identity over secrets, Git from day one.
Open → -
Master Sequence Page: One Page That Sequences Every Instruction Page
markdown from The Context Harness and the Execution Harness: Two Harnesses, One Loop
The page I hand Claude at the start of every session on a big project: current phase, status legend, instruction page sequence, write-back protocol, benchmarks.
Open → -
Fabric Data Agent Instructions: Two Sub-Agents on One Semantic Model
markdown from A Data Agent Should Be a Sub-Agent
Complete instructions for two Fabric data agents sharing one semantic model, each written as a sub-agent that Foundry or Copilot calls for one job.
Open → -
Data Contract Watchdog Brief
markdown from Governance Pillars 3 and 4: Write the Contract Before You Hire the Watchdog
The one-page scope for an agent that checks your lakehouse against signed data contracts every morning and notifies owners by name.
Open → -
CoE Documentation Sweep Brief
markdown from AI Made Your Center of Excellence Bigger. Did It Get Better?
The level-one instruction page for an agentic Center of Excellence: document every column, measure, and lineage path over MCP, and say UNKNOWN instead of guessing.
Open →
Agent Skills & Definitions
Skill files and agent definitions, front matter included: what triggers them, what they read first, and the guardrails that keep them in their lane.
-
Grill the Fabric Plan
markdown from A skill.md Is Not Wisdom. So Make the Agent Grill You.
A one-page skill that interrogates a Fabric build before anything gets created, by naming the design tree and delegating the interview.
Open → -
Project Instruction Writer
markdown from Build the Thing That Creates the Thing
The generator behind "update the instructions for this project": turns meeting notes and hub context into the instructions page a build session runs on.
Open → -
Skill Converter: Rewrite a Claude Skill as a Notion AI Skill
markdown from Build the Thing That Creates the Thing
The full converting-skills-to-notion skill I run. Turns a Claude skill into one Notion page that works as an AI Skill and as a context page Claude reads back later.
Open → -
Department Vocabulary Skill
markdown from AI Made Your Center of Excellence Bigger. Did It Get Better?
One skill file per business function, sourced from that function: the words they use, the words they refuse, their real process, and a named owner.
Open → -
Fabric Advisor Orchestrator Agent
markdown from Fabric Task Flow Studio: From One Sentence to a Deployed Fabric Architecture
The agent definition that routes a seven-phase Fabric pipeline to one skill per phase, and turns the sign-off gate into a plain-language approval instead of a decision table.
Open → -
Fabric Prompt Builder Skill
markdown from Fabric Task Flow Studio: From One Sentence to a Deployed Fabric Architecture
The agent skill that turns Notion pages, meeting notes, and mapping docs into a single high-signal intake prompt for the Fabric task-flows pipeline.
Open →
Fabric Recon & Guardrails
What an agent reads before it touches a tenant: read-only survey briefs and dated verified-state blocks.
-
Fabric Reconnaissance Brief and Verified Tenant State
markdown from Don't Let Your Agent Touch Fabric Until It Reports Back
The read-only survey an agent runs before it touches a Fabric workspace, plus the dated state block that tells it how your tenant is misleading.
Open →
DAX & Validation Queries
Measure patterns and the DAX acceptance tests that decide whether model work counts as done.
-
DAX Acceptance Tests for an MCP Modeling Session
dax from From Notion Page to Power BI Model: Inside an MCP Execution Session
The EVALUATE queries an agent runs before it reports model work as done, plus the smart anchor pattern and how a documented deviation should read.
Open →
Fabric Notebook Patterns
PySpark scaffolding and validation cells: the NB_TEST switch, referential integrity checks, domain asserts.
-
The Fabric Estate Audit: What Central BI Should Be Able to Answer Without a Meeting
python from Central BI Is a Tiger Team Now
A read-only notebook that inventories every workspace, item, domain, and admin in the tenant, then asserts the governance rules you claim to have.
Open → -
Fabric Notebook Test Harness: NB_TEST and the Validation Cell
python from Don't Let Your Agent Touch Fabric Until It Reports Back
The PySpark header that lets a notebook prove itself locally before it touches OneLake, and the asserts that gate every run after.
Open →
Session Kickoff Prompts
The short lines you actually type, once the context engineering already happened somewhere else.
-
MCP Session Kickoff and the DAX Autonomy Directive
markdown from From Notion Page to Power BI Model: Inside an MCP Execution Session
The one line I type to start a modeling session, and the standing directive that turns the agent from a template-follower into a modeler.
Open →
Playbooks & Checklists
Meeting scripts, scorecards, and rollout checklists for the human side of this work.
-
Skill Library Playbook: How a BI Team Owns Its Agentic Knowledge
markdown from Agentic Distributed Wisdom: Why Your Team's AI Skills Can't Live in One Head
The one-page playbook a BI team keeps next to its shared skill library: where skills live, how they're added, reviewed, and maintained, and the monthly check that says the knowledge is distributed.
Open → -
Tiger Team Engagement Charter: Discovery, Accountability, and the Exit
markdown from Central BI Is a Tiger Team Now
The three documents that keep a floating central BI team from quietly becoming a department again: the discovery brief, the domain accountability table, and the charter with exit criteria.
Open → -
The Pipeline Interrogation Gate
markdown from Don't Hire a Senior to QA the Agent
The questions whoever ran the agent answers with the agent closed, before a senior opens the notebook at all.
Open → -
AI Tool Intake Test
markdown from You Don't Have an AI Strategy. You Have AI Theater.
Five questions and a removal trigger that decide whether a new skill, MCP server, or harness gets near client work.
Open → -
Data Contract Template
markdown from Governance Pillars 3 and 4: Write the Contract Before You Hire the Watchdog
The agreement humans sign before any agent gets to enforce anything: allowed values, definitions, timing, and the violation mode picked in advance.
Open → -
The DAX Readiness Gate
markdown from Agents Raise the Floor and Lower the Ceiling
The three gates, four trap measures, and pass criteria I use to decide whether someone is ready for agentic DAX tooling.
Open →