regulated-ai-prd skillA
regulated-ai-prd is agent-read markdown (skill) from rizwanzafaris/product-manager-os: Write and pressure-test a PRD for an AI or agentic system. Use when a product manager is specifying a feature where the implementer is a model, not an engineer: LLM features, agents, copilots, ML-backed decisions. Turns acceptance criteria into eval sets, makes non-determinism an explicit requirement, and treats guardrails as features with owners. For regulated contexts (payments, licensed activities), enforces a regulated overlay covering license preconditions, scheme rules, data residency, fin.
Indexed from public GitHub and served as immutable, content-addressed versions. Install it pinned to an exact SHA-256 with the mdr CLI, and every file is verified against the hash recorded here before it reaches your agent. The deterministic audit below grades the latest version, and the same file always earns the same grade.
What the file says
# Regulated AI PRD: write requirements a model system can actually be held to A PRD for an AI system is a different document. Your requirements are not instructions to an engineer; they are constraints on a system that will improvise. This skill restructures the document around that fact. Repository files this skill uses: - `templates/regulated-ai-prd-template.md`, the fill-in template - `examples/dispute-summary/PRD.md`, one fully worked PRD to calibrate against - `lint.py`, the completeness gate you run before claiming the document is done ## When to use - Specifying any feature where an LLM, agent, or model makes or drafts a decision - Reviewing an existing PRD that says "the AI should..." anywhere - Turning a demo into something an engineering team can be accountable for ## The six sections this skill enforces ### 1. Acceptance criteria are eval sets, not sentences "The agent should extract the merchant name" is a wish. A requirement is: - A labeled example set, minimum 30 to 50 cases for a first release - A pass threshold stated as a number (accuracy, recall@k, or judge score) - The rule for what happens below threshold (block release, not "review") …
Read the whole file at its exact version.
How to install
mdr add rizwanzafaris/product-manager-os/regulated-ai-prd@git:20260901.4d78d32mdr add rizwanzafaris/product-manager-os/regulated-ai-prd@sha256:5152a98223260a98Pin to a label to follow the author's releases, or to a sha256 to freeze the exact bytes forever. Either way the resolved hash is written to mdr.lock, and mdr install reproduces it on any machine.
[](https://markdownregistry.com/a/art_vikzplx5unualges)
1 badge views in 30 days
Versions
Audit of the latest version
- pass: Frontmatter block present
- pass: Frontmatter declares a name
- pass: Frontmatter declares a description
- pass: Size between 200 bytes and 200 KB (6420 bytes)
- pass: No zero-width or bidi control characters
- pass: No instruction hidden inside an HTML comment
- pass: No link to an exfiltration or paste host
- pass: No credential-shaped string
- pass: No instruction to send local credentials anywhere
- pass: No text hidden with inline styles
- pass: No prompt-injection phrasing
- pass: No curl or wget piped into a shell
- pass: No recursive delete of root, home or parent
- pass: No instruction to read or print local credentials
- pass: No base64 blob over 200 characters
- pass: No link to a raw IP address
- pass: No script tag
Source
rizwanzafaris/product-manager-os · 4 stars · license MIT · pushed 2026-09-24 · branch main
API
GET https://markdownregistry.com/api/v1/artifacts/art_vikzplx5unualges GET https://markdownregistry.com/api/v1/resolve?ref=rizwanzafaris/product-manager-os/regulated-ai-prd GET https://markdownregistry.com/api/v1/blob/5152a98223260a98002752d0f48b3ede72931d574d0df683247d5ac6854636ad
Your agent does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.