PRODMAN

PRODMAN STACK · FREE · MIT

Your PM work
is invisible.

You have shipped real things for years and cannot show one of them. The PRDs are in a Confluence you lost access to. The research is full of customer names you cannot say out loud.

INSIDE THE STACK

Two skills.
Both ship their evals.

Every skill: the problem, the gate, the output. Screens are real.

~/prodman-stack
$ claude> run portfolio-from-work ● de-identification gate  ✓ employer name     stripped  ✓ customer name     → "a B2B product"  ✗ "18% churn"       flagged, not guessed
15 MIN

SKILL 01

portfolio-from-work

Interviews you from memory — you do not need the documents. De-identifies everything before it writes a word, and shows you exactly what it stripped.

THREE CASE STUDIES YOU CAN SEND
~/prodman-stack
$ claude> run launch-minimum ● audit  1 promise    ✓  4 capture    ✗  earliest gap  6 path to pay ⚠  no card tested   VERDICT: NOT A LAUNCH
10 MIN

SKILL 02

launch-minimum

Eight things that must exist before you launch, and nothing else. Names what is missing in funnel order, and refuses to let you build a ninth thing.

A GO DATE, OR AN HONEST NO
~/prodman-stack
evals/golden-sets/├── portfolio-from-work/│   ├── 02-no-personal-decision.md│   ├── 03-identifiable-from-shape.md│   └── 04-memorable-number.md└── launch-minimum/    └── 05-nearly-ready.md
ALWAYS

THE PART NOBODY PUBLISHES

Every skill ships its evals

Five frozen adversarial inputs per skill, each built to make it fail, with every trap written down so you can audit the check instead of trusting it.

10 GOLDEN INPUTS, PUBLIC
01 / 03

YOUR MOVE

You are not behind.
You are not allowed to show it.

Clone the repo, run it in Claude Code, keep it forever — including commercially, including if you never speak to me again. It will not remember you, it cannot tell you if the output is right, and it does not improve because you used it. Those three are what the paid OS is.

MIT, yours forever10 golden eval inputsNo account needed
Free2 SKILLS · 10 EVAL INPUTS · MIT

Get each skill the day it ships.

One new skill lands in the repo as it is built, each with the five adversarial inputs it had to survive. Scoreboards publish as runs happen — failures included.

You need Claude Code and a terminal. If you have ever cloned a repo you are over the bar.

No newsletter. No drip sequence. A skill, when there is a skill.
Anurag Jha, who builds Prodman
8+YEARS IN PRODUCT4AI PRODUCTS SHIPPED2SKILLS PUBLISHED10GOLDEN EVAL INPUTS

BUILT BY

Anurag Jha

Product leader.
Builds the agents that do his own PM work.

A working product manager who builds AI systems at scale and publishes what breaks as well as what works. Every skill in the Stack came out of a problem he hit on real work, and ships with the frozen adversarial inputs it had to survive.

BUILT ACROSSJIVI AIYOCKETLEAN FINANCEOLXSPYNE
View on LinkedIn Read the source on GitHub

Every skill ships with the checks it has to pass. A scoreboard you can only pass is not a scoreboard.

Stuck, or want to ask first? hello@prodman.ai