Infinum · Engineering · The Stack

Infinum
AI Stack.

A workflow, a spec engine, a set of skills, and an in-house agent — packaged together and installed per repo with one command — plus an org-wide tracker for how it's all being adopted.

More details, the full skill catalogue, and install commands are available at github.com/infinum/ai.

Why a stack: building with AI without a shared setup produces inconsistent, hard-to-maintain results. After two years of practice we've packaged that setup — a workflow, a spec engine, skills, and an agent installed per repo, plus an org-wide view of how it's used — living alongside the code it works on.
What's in the stack
01
PSDD ↗
Pragmatic Spec Driven Development — the 4-step workflow. Depth scales with the work.
02
Superpowers
The spec-driven engine inside PSDD. Brainstorm, plan, TDD, self-review — all as skills.
03
Skills marketplace
PRD authoring, security audits, UI validation, local PR review — each one a separate plugin.
04
Phantom
Our Slack-native AI agent. Drops a fix in your repo from a Slack thread, runs your validation, opens the PR.
05
AI Stats
Internal overview of AI-tool usage, cost, and adoption across teams and skills.
One-time setup
$
Install the stack with one command
Idempotent — re-run any time to update rules or pick up new plugins

A single setup script registers the Infinum marketplace in Claude Code, lays down team rules and personalization stubs, and walks you through an interactive prompt to install the skills you want. Re-runnable; safe.

$ pnpm dlx github:infinum/ai

How and why to use pnpm?

Project-level over user-level. When configuration lives in the repo, improvements travel with the project — not just with the engineer. Also installable via the pnpm dlx --allow-build=skills-cli skills CLI for Cursor, Copilot, and Firebender.

The five parts, in detail
01
PSDD — the workflow ↗
Pragmatic Spec Driven Development
Workflow

Infinum's 4-step workflow for shipping features with AI. The same four steps apply to every task — depth scales with complexity and risk.

→Understand. Remove uncertainty before anything else. PRD, technical brief, or short prompt — depth scales with risk.
→Design. Agree on the technical direction. No code yet. Ask the model how it would build it.
→Plan. Turn the agreed design into an ordered, reviewable plan. Last clear checkpoint before code runs.
→Build. The agent runs. Because the upfront work is done, it stays aligned. Corrections are fast and targeted.
Three lanes: small / low-risk compresses all four steps into one prompt; purely technical work skips the PRD for a tight technical brief; large or high-risk work gets every step in full. Read the full guide at infinum.github.io/ai/psdd.
02
Superpowers — the spec engine
Brainstorm → spec → plan → code → review, all as skills
Engine

PSDD is the workflow. Superpowers is the engine inside Steps 2 and 3 — the spec-driven development plugin we adopt as the default. It treats spec, plan, and review as first-class skills, not afterthoughts.

→Brainstorming and exploration before any plan exists.
→Written specs reviewed by independent subagents — parallel passes to catch ambiguities.
→Implementation plans with phases and verification gates — not "go write the code."
→TDD as a default, not an option. Self-review before "done."
Superpowers is open source by @obra. It's the recommended default in our setup, but the stack is agent-agnostic — bring your own equivalent if you prefer.
03
Skills — per-repo plugins
Install only what the project needs
Skills

Each skill is a separate installable plugin. Engineers add only what they need, updates ship per-skill, and the configuration lives next to the repo it works on. Improvements travel with the project — not just with the engineer. A selection of what we ship today:

create-prd
Generates a structured PRD in Markdown. Guided flow with clarifying questions, iterative refinement, and a consistent structure — goals, non-goals, user stories, acceptance criteria, constraints.
claude-setup-audit
Security audit of an engineer's AI tooling. Scans the local Claude Code installation for vulnerabilities, malicious code, prompt injection, and risks across every extensibility point. We audit our tools the way we audit our codebases.
ui-validation
Figma-vs-implementation diff before QA gets involved. Snapshot tests plus a multimodal visual comparison loop — implement with shared components, snapshot, fetch Figma, compare, fix or stop-and-ask. Framework-agnostic.
pr-review-code-simplicity
Local agentic PR review before you push. Reviews against the six laws of software design from Code Simplicity (Max Kanat-Alexander). Findings focus on long-term maintenance cost, not style.
package-security-check
Supply-chain audit of a package before you update it. Layered checks across GitHub source, registry metadata, install scripts, known CVEs, maintainer changes, and the version diff.
This is a selection, not the full catalogue. The marketplace grows continuously. Current list and install commands live in the repo at github.com/infinum/ai.
04
Phantom — the Slack-native agent
Fix, build, review, explain — from a Slack thread
Agent

An AI agent we built ourselves and run on real production projects. Open a Slack thread, drop the context, mention @Phantom, and an isolated, sandboxed agent clones the repo, reads the thread, applies the change, runs your validation, and opens a pull request — assigned to a human owner.

@Phantom fix @Phantom build @Phantom review @Phantom explain @Phantom vuln-fix
→Sandboxed Docker containers per job. Per-stack images for Android, Flutter, JVM, .NET, MAUI, Ruby, and a base image. No ambient repo access, no shared state.
→Self-registers by DM. No YAML edits, no setup PR. Open a chat with Phantom, say "set up a project," and it walks you through.
→Human ownership on every PR. Phantom opens; humans review and merge.
01 thread
QA or dev posts the bug, context, screenshots, Figma link.
02 mention
@Phantom fix — the agent reads the whole thread.
03 work
Clones, fixes, runs lint / tests / build in a sandbox.
04 PR
Opens a pull request, reports back in the thread.
05
AI Stats — the measurement layer
Who's adopting AI, on what, at what cost
Insights

The other four help you build with AI. AI Stats measures how much it's used — a web app that aggregates every employee's AI-tool usage into company-wide adoption and spend.

→One schema, many tools. Cursor, Firebender, Claude Code, and Copilot — normalized into a single database.
→Breakdowns by user, tool, model, team, skill, and month. Filter to the slice you care about.
→Skill-level adoption. Per-skill Claude Code usage — tied back to the skills in 03.
→The whole picture, not just spend. Usage and cost are paired with actual code changes from GitHub / GitLab / Bitbucket — so adoption is measured against what shipped, not tokens alone.
→Role-based access. Viewer, Contributor, Manager, Owner.
Glossary
Terms used in the stack
The Stack
The packaged combination of workflow, spec engine, skills, and agent that Infinum engineers install per-repo, plus AI Stats for measuring adoption across the org. Described in this document.
PSDD
Pragmatic Spec Driven Development — Infinum's 4-step workflow for building with AI. Full guide at infinum.github.io/ai/psdd.
Superpowers
Open-source spec-driven development plugin by @obra, adopted as our default. Powers Spec → Plan → Code with brainstorming, planning, TDD, and self-review skills.
Skill
A focused, reusable instruction set installed as a plugin. Defined in a SKILL.md file. Each Infinum skill ships as its own plugin so engineers install only what they need.
Phantom
Infinum's in-house Slack-native AI agent. Runs jobs in isolated Docker containers; opens PRs assigned to a human owner. Repo: infinum/ai-phantom.
AI Stats
Infinum's AI-tool usage tracker. Aggregates every employee's usage into adoption and spend by user, tool, model, team, skill, and month.