Skip to content

Flagship

Flagship helps teams run product experiments faster and with less risk by using feature flags and cohort analysis.

Code Video Cerebral Valley

Core trick

An agent skill serves feature-flag variants, reads PostHog cohort metrics on a daily loop and keeps the best-performing variant through human-reviewed PRs.

Tags

dev toolsagents

Description

Flagship helps teams run product experiments faster and with less risk by using feature flags and cohort analysis. It measures how different user cohorts respond to each variant and proposes code changes based on the results. Humans still review and merge every PR.
TwillAI/flagship★ 10 forksShellCreated Feb 5, 2026 · last push Feb 6, 2026

README excerpt

# Flagship

> Disclaimer: Flagship is experimental software. APIs, workflows, and behavior may change without notice, and it is not recommended for production-critical use without independent validation.

Flagship is an **Agent Skill** (`skills/flagship/SKILL.md`) that helps teams run product experiments faster and with less risk by using feature flags and cohort analysis.
Given a budget, an agent using this skill runs the experiment, analyzes cohort results, and keeps the option that performs best.

## High-Level Behavior

```mermaid
flowchart TD
  A["Define experiment (objective, KPI, budget)"] --> B["Create initial change and open PR"]
  B --> C["Human review and merge"]
  C --> D["Runtime serves variants and records events in PostHog"]
  D --> E["Daily analyze loop runs metrics + policy gates"]
  E --> F{"Decision"}
  F -->|"HOLD / SHIP"| D
  F -->|"ITERATE"| G["Generate follow-up PR and merge"]
  G --> D
```

## What It Does

Flagship helps you test product ideas without ad-hoc scripts. It combines cohort data from PostHog (via MCP), deterministic safety gates, and Codex-generated treatment updates. The result is a repeatable loop: measure, decide, propose, review, and merge. You keep control through budgets, guardrails, and normal PR review.

## Quick Start (5 minutes)

1. Create a GitHub Environment named `flagship`.
2. Add secrets: `OPENAI_API_KEY`, `POSTHOG_API_KEY`, `POSTHOG_MCP_URL`.
3. Create an experiment in Codex:

   ```text
   Use the flagship skill and create

Team