This means agentic integration for multiple products across multiple frameworks. This also means support for additional “skills” like migrations and troubleshooting. Ultimately, the Wizard closes the gap between curiosity and real usage. It's an agent that delivers working PostHog solutions within a user’s codebase in minutes.

This means our content like docs, handbook, and blog posts play a central role in context engineering. By structuring our content into composable and reusable context, like skills, we can build agentic workflows that show how the entire product suite works together in practice. Success means developers see our content platform as both a well-organized knowledge base for readers and a context-rich ecosystem for agents.

This means each product engineering team writes and maintains their own product docs at a strong baseline level of quality by using the templates, tools, and processes we’ve put in place. We partner with teams to review, improve, or overhaul their docs when needed. Success means keeping current product docs updated and launching several new products with docs, without slowing down.

Driver: Vincent GeVincent Ge

Docs & Wizard Team

Docs & Wizard Team

Motivation: Make the wizard harness more flexible and modular so it can run a variety of agentic workflows. The wizard should be able to execute any skill provided by the context mill.

What we'll ship:

• Wizard harness that is agnostic to any skill it runs
• Wizard harness that can queue arbitrary additional tasks
• Swappable system prompts to bootstrap different workflows
• Composable TUI UX to handle different workflows and screens

We'll know we're successful when:

• Anyone at PostHog can deliver an entirely new workflow on the wizard with a context mill
• Wizard has skills for every PostHog product
• Wizard has skills for non-integration use cases like migration, troubleshooting, cost cutting, etc.

Make it weird:

• PostHog has many fun people with fun opinions. Distill them into skills, so they can roast or give constructive advice on your code base while the Wizard runs.

Driver: Vincent GeVincent Ge Danilo CamposDanilo Campos

Motivation: The wizard is a powerful integration opportunity for several teams at PostHog. We should expose just enough data and interfaces so other teams can connect to the wizard.

What we'll ship:

• Headless mode for the wizard
• HTTP calls that emit wizard data and task status
• Uploaded event schemas and data artifacts to the user's PostHog project after each wizard run

We'll know we're successful when:

• 
  Website Team
  
  Website Team can build a sandboxed wizard run on the website
• 
  Growth Team
  
  Growth Team can build a wizard-synced in-app onboarding page
• self-driving can generate signals from wizard runs
• 
  PostHog Code Team
  
  PostHog Code Team can run the wizard inside PostHog Code

Make it weird:

• While the Wizard runs, both Wizard and in-app onboarding should play reels style demos and offer games to compete for user attention. HogToks are in, TikToks begone.

Driver: Danilo CamposDanilo Campos

Motivation: We must up our own game. Our codebases don't have a good agent onboarding experience. First-time wizard contributors are confused, unaware of best practices, and opening convoluted PRs.

What we'll ship:

• Repo-level skills to support full-stack wizard contributions across wizard, context mill, and workbench
• Ongoing maintenance of skills, agents|claude.md, and README files to prevent drift as the project evolves
• Codified context engineering principles and practices

We'll know we're successful when:

• We get regular wizard contributions that are correct and easy to merge
• Agents produce sublime wizard-compatible architectures without extensive prompting
• Newcomers have a fun and delightful first experience working with the wizard

Make it weird:

• It's Danilo. He'll find a way.

Driver: Edwin LimEdwin Lim Danilo CamposDanilo Campos

Motivation: Agents should be 10x-ing the business, not just the codebase. It's time to unleash the GTM and commercial teams. Let them experiment with, create, and ship killer software to customers using skills, context mill, and the wizard. SKILL CREATION FOR EVERYONE.

What we'll ship:

• Skill creation interface, tools, and resources
• Skill maintenance and collaboration systems with versioning
• Skill catalogs and distribution through posthog.com, GitHub, Claude plugins, wizard, etc.

We'll know we're successful when:

• Teams are using skills + wizard to help close deals, nail demos and trials, and get customers excited about PostHog
• Folks are spontaneously creating and shipping spec-compliant skills to solve internal and customer problems
• We turn the company handbook into a skillbook
• Docs are viewed as skills

Make it weird:

• Incorporate an IRL magic show or scripted component for customer demos as the wizard runs skills

Owner: Sarah SandersSarah Sanders

Motivation: If an agent goes rogue, who can stop it? Another agent of course. We're going full Terminator 2 and sending Sarah Connor Sanders to build and join forces with a security-focused agent to lock down the wizard and its context as potential attack vectors.

What we'll ship:

• A standalone security service (codename: warlock) that scans agent context and tool input/output for malicious content
• An agent harness to run the warlock and analyze its findings
• Alerting if threats are detected
• Security best practices for context and agent engineering

We'll know we're successful when:

• Nothing happens, bc nobody can use the wizard to do bad things 🤞
• We find malicious threats that we know exist in public distros like skills.sh
• 
  Security Team
  
  Security Team has blessed our approach

Make it weird:

• Warlock's Most Wanted - work with the art team to create old-west style bounty posters for the sketchiest skills found

Driver: Edwin LimEdwin Lim

Motivation: Agents are kind of a magical content experience. People’s eyes are glued to the TUI as the wizard runs. We can put on a great show: part demo, part onboarding, part spectacle. Let’s get weird.

What we'll ship:

• A more robust text adventure-style rendering system for wizard "Learn" content
• Real-time streaming of wizard discoveries about the user's app for personalized storytelling
• "Winning with PostHog" adaptation
• Refresh website /docs page to be wizard-first

We'll know we're successful when:

• Users mention the wizard's content in #brand-mentions
• Customer-facing teams consider it a strong demo and learning resource
• It sparks joy and curiosity

Make it weird:

• "The Life of a Data Point", a text adventure game about an anthropomorphized data point, born in the user's app, finds its way to becoming a product

Owner: Sarah SandersSarah Sanders Vincent GeVincent Ge

Motivation: The InKeep Content Writer agent was a hit. But how do we keep the momentum going? We should improve the DX even more to make it even easier to ship high-quality docs. That's how we protect and improve our knowledge base as a reliable source of context for agents.

What we'll ship:

• A scheduled audit agent to review deterministic sections like SDK and API references
• Our own standard for what "high AI context quality" means in our docs, and a way to measure and test it
• Solve the lag problem between website builds and merging InKeep PRs
• Culling and replacing screenshots with prompts
• Improve how we trigger InKeep Content Writer from Slack
• A leaderboard of the best docs contributors!

We'll know we're successful when:

• We break the record for most docs PRs merged in a quarter
• Product eng teams are actively trying to improve their docs quality score
• Engineers enjoy using docs agents like InKeep even more and feel more ownership
• Our agents perform better because context quality is improving

Make it weird:

• Make the leaderboard an actual leveling system. Earn XP for PRs merged, quality scores, streaks. Titles evolve over time, you can pick a class, and it tracks stats differently.

Driver:

Docs & Wizard Team

Docs & Wizard Team

Motivation: Tell the wizard story, a lot. Sharing what we've learned with others is how we become experts.

What we'll ship:

• Turn our projects into published pieces of content (blog, conference talk, social post, podcast interview, etc.)

We'll know we're successful when:

• People learn from and cite our work
• We're seen as leaders in context engineering

Make it weird:

• We have one of the weirdest jobs in tech. Just share what we're doing.