A gated PRD lifecycle for Claude Code: stage skills from problem framing to growth, a Markdown knowledge graph, hooks and readiness scores.
by mattgierhart· Project managementUnclaimed · Yours? Claim it
4 images
About
PRD-Driven Context Engineering is Matt Gierhart's community methodology for building products with AI agents that remember, packaged as a Claude Code plugin. It keeps your product's decisions in plain Markdown, each with a typed ID (BR-004, UJ-101, API-045), so your agent loads just the context a task needs. And it moves the PRD through gated stages, from v0.1 to v1.0, instead of building everything at once. What it adds - /prd-ce:init seeds a fresh repo with PRD.md, Source-of-Truth files in SoT/, EPIC templates and the ID registry. - Stage skills for each PRD version: problem framing, market definition, pricing, personas and user journeys, then risk review, architecture, the build loop and test planning, then release, launch and growth. - Launch and growth skills encode named frameworks: April Dunford positioning, Alex Hormozi offers, Geoffrey Moore's chasm, Teresa Torres's continuous discovery and Mom Test interviews. - Operator skills (ghm-*) for gate checks, building SoT entries, registering IDs, harvesting insights and syncing status. - Four role agents, horizon (strategy), studio (design), devlab (build) and metro (ops), each with its own memory file. - Readiness scoring: a script scores SoT files, EPICs and the PRD stage, and says whether you're ready to advance and what to fix first. - Hooks that give your agent the read order at the start of a session, check context density on each prompt, and remind it to update SoT after code changes. How to install it In Claude Code, run /plugin marketplace add mattgierhart/PRD-driven-context-engineering, then /plugin install prd-ce@prd-ce-methodology, and run /prd-ce:init in a fresh repo. You can also fork the repo as a template, or run its install.sh against an existing repo. Good to know - It works on files in your repo, with no server, account or API key. Its scripts need Python 3 and bash. - Its hooks run in every Claude Code session while the plugin is enabled. In a project with an epics/ folder, it asks before code edits when no EPIC is in progress. - /prd-ce:init seeds new repos only for now; an existing PRD.md or SoT/ is left as it is. - It's a community project by Matt Gierhart, open source under the MIT licence.
Once it’s added
“Let's frame the problem.”
“Are we ready to move the PRD to v0.2?”
“Map the core user journeys for our app.”
“Scope the next EPIC from the PRD.”
“Write our positioning using April Dunford's method.”
“Plan Mom Test interviews with our early customers.”
Before you add it
Reads only
PRD.md, SoT files and EPICs, loading only the IDs a task needs · How ready each SoT file, EPIC and PRD stage is
Makes changes · your agent can ask first
PRD.md, stage by stage from v0.1 to v1.0 · SoT entries with typed IDs, such as BR-, UJ- and API- · EPICs, agent memory files and lessons learned
Each with its own page
From people who use it
Step by step
No reviews yet. Used PRD-Driven Context Engineering? Say how it went.
Write a reviewRuns on your machine.
Runs on your machine.
/plugin marketplace add mattgierhart/PRD-driven-context-engineering /plugin install prd-ce@prd-ce-methodology
Scanned and passed · checked 6 h ago
Plugins people install alongside PRD-Driven Context Engineering.
Nothing new reaches your agent until a person has looked at it.
Scanned on every change, and again when our rules improve.
Claude CodeSupported
OpenClawSupportedTell us and a person on the trust team will look at it. Or email [email protected].
Nobody has shared how they use PRD-Driven Context Engineering yet.
Security scan
Passed
Safe to install in every agent it supports. Claude plugins.
Plugins people install alongside PRD-Driven Context Engineering.
Linear's plugin from Anthropic's directory: adds Linear's hosted MCP server to find, create and update issues, projects and comments.
A GitHub issue resolver, a standup-notes command and a developer-experience subagent, for teams working with a coding agent.
A community skill that has your agent write INVEST user stories with acceptance criteria, split epics and plan sprints from your velocity.
