Skip to content

raziiabraham/pm-mts

Repository files navigation

PM MTS Companion

The AI-guided practice companion to PM Is Now Another Member of Technical Staff by Razii Abraham.

You have done the reading. Now let the repository respond.

Open or attach this repository in an AI coding agent and say:

Guide me through Chapter 2.

Or use the repository-native invocation for your agent:

  • Codex: $pm-mts-guide Chapter 2
  • Claude Code: /pm-mts 2

The agent should reconstruct the API request and responses in the conversation, ask for your product judgment, run the safe contract check itself, explain the output, then interview you through an API State Worksheet. You should not have to hunt through files, copy terminal commands, or fill in a blank workbook alone.

For another chapter, change the number. For your own product, add one sentence of context:

Apply Chapter 8 to our checkout recovery flow. Interview me one question at a time and challenge missing states.

See AI_GUIDE.md for demonstration, application, and experiment prompts, including compact command examples.

What happens in an AI-guided session

  1. The example appears in chat. The agent reads bounded repository sources and renders the relevant request, query, event, flow, evaluation case, or decision table.
  2. You make a prediction or judgment. The agent asks one useful question before explaining the supplied result.
  3. The agent runs safe checks. When a no-key fixture exists, the agent executes it, shows the salient output, and explains what it proves.
  4. You apply the technique. The agent interviews you one question at a time, gives specific feedback, and maintains the workbook record in the conversation.
  5. The result gets challenged. The agent tests one failure state, counterexample, or unsupported assumption.
  6. You leave with an artifact. The completed draft, unresolved questions, evidence boundary, reviewers, and next action are rendered in chat. Saving a file is optional.

Repository files remain available for provenance, deeper inspection, reproducible evidence, and reuse. They are not the default user interface.

Choose a practice session

I want to practise… Say this What the agent brings into chat
API product states “Guide me through Chapter 2.” Host, method, path, request, responses, UI decisions, and a live contract check
SQL product reasoning “Guide me through Chapter 5.” Exact query, result table, grain and denominator walkthrough, and a live SQLite run
Product analytics “Guide me through Chapter 6.” Tracking decision, valid and invalid events, validator output, and a guided tracking plan
Experiment judgment “Guide me through Chapter 7.” Decision rule, result scenarios, interpretation feedback, and a pre-registration draft
AI evaluation “Guide me through Chapter 13.” Held-out cases, candidate outputs, release gate, evaluator output, and a scorecard draft
AI calibration “Guide me through Chapter 14.” Preserved failure, version chain, regression logic, check output, and a calibration record
Agent design “Guide me through Chapter 15.” Goal, tools, memory, budgets, stops, authority, challenges, and an agent charter
Repository orientation “Guide me through Chapter 17.” Inspection map, classified findings, contradictions, unknowns, and a bounded check when available
Any chapter “Apply Chapter N to [situation].” A progressive interview, feedback, completed record, evidence boundary, and next action

The Companion Map covers all 20 chapters. The quick path remains Chapters 1, 8, 13, 16, and 20; add Chapter 15 when the product itself contains an agent. The capstone path begins with one evidence row in Chapter 18 and carries it through prototype and delivery in Chapters 19–20.

Three modes

  • Demonstrate: reconstruct a supplied example, execute its bounded evidence, and explain the product meaning.
  • Apply: interview you and produce the relevant decision artifact in chat.
  • Experiment: record a prediction, make one reversible change, show the diff, run the narrow check, and compare the result.

If you do not choose, the agent starts with a demonstration and then offers to apply the technique to your product.

What is included

Area Source material and executable evidence
companion/chapters/ Conversation routes for every chapter
companion/templates/ 26 chapter-aligned schemas used by guided interviews
companion/api/ A complete fictional API contract and deterministic validator
companion/sql/ The book’s exact SQL and open-licensed Sakila dataset
companion/analytics/ Tracking plan, valid and invalid event envelopes, and validator
companion/ai_evaluation/ Two constructed AI-output versions, held-out cases, rubric, and evaluator
companion/calibration/ A preserved failure, diagnosis, change, regression evidence, and release chain
companion/repo_orientation/ A bounded public-repository orientation exercise
companion/feedback_instrument/ A synthetic labelled-feedback dataset with edge cases
companion/figures/ Every book diagram as PDF, PNG, and SVG

If your agent does not have the repository yet

Give it the repository URL—https://github.com/raziiabraham/pm-mts—or use your agent’s GitHub/open-folder flow. Ask the agent to obtain or open the repository for you. Manual clone and download options remain in GETTING_STARTED.md as a fallback, not the default lesson.

The supplied path needs no model API key, cloud account, production credential, or customer data. A terminal-capable agent can run the checks with Python 3.11+, SQLite 3.40+, and a POSIX-compatible shell.

Manual and maintainer verification

If you intentionally want to reproduce every check yourself:

./scripts/validate_companion.sh

This is an evidence and maintenance path, not the reader’s starting point.

Downloadable materials

Evidence and data boundary

A green check proves only the named assertions about the supplied teaching fixture. It does not prove that a production API, query, AI feature, or product decision is correct. The reader and relevant specialists remain responsible for interpretation, evidence, permissions, and real-world verification.

Named companies and people in constructed runnable fixtures are fictional, and all feedback reviews are synthetic. Chapter 5 vendors the open-licensed generated Sakila sample data. Chapter 17 intentionally inspects the author’s public Noted repository at a cited commit. This repository contains no employer code, credentials, customer data, or confidential artifacts.

Copyright © 2026 Razii Abraham. All rights reserved.

About

Companion workbook, templates, and no-key exercises for PM Is Now Another Member of Technical Staff

Topics

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages