ce-dogfood

Runs diff-scoped browser QA on a branch, fixing small breakages and writing a dogfood report.

24.7k|2.0k|Updated Oct 9, 2025
One-click install
npx skills add https://github.com/EveryInc/compound-engineering-plugin --skill ce-dogfood
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: ce-dogfood
Source: https://github.com/EveryInc/compound-engineering-plugin/tree/main/skills/ce-dogfood
Command: npx skills add https://github.com/EveryInc/compound-engineering-plugin --skill ce-dogfood

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Manually verifying that every user-visible change on a feature branch actually works end-to-end in a real browser is slow and easy to skip, so regressions and UX friction ship unnoticed.

Core Features & Use Cases

  • Diff-scoped browser testing: Maps the user flows a branch introduces versus the trunk, derives a test matrix, and drives a real browser through each journey via the agent-browser CLI.
  • Autonomous fix loop: Fixes small, low-risk breakages with regression tests and commits, while escalating large or ambiguous changes to a "Decisions for a human" report section.
  • Persona-based experience review: Judges each flow against product personas, logging paper cuts and console errors alongside functional results.
  • Use Case: Before merging a pull request that adds a new checkout flow, invoke the skill on the PR to have every scenario exercised in a browser, small bugs fixed and committed, and a durable dogfood report written to the repo.

Quick Start

Ask the agent to dogfood the current branch or a specific PR, for example by invoking ce-dogfood with a PR number to run browser QA and produce the report.

Frequently Asked Questions about ce-dogfood

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run automated browser QA on a pull request before merging?

Invoke ce-dogfood with the PR number. It checks out or isolates the PR, maps the user flows from the diff, drives a real browser through each scenario with agent-browser, fixes small bugs with regression tests, and writes a dogfood report with a readiness verdict.

What is diff-scoped browser testing?

Diff-scoped testing exercises only the user-visible changes a branch introduces versus the trunk, rather than exploring the whole app. The skill derives the diff, maps affected flows as Mermaid diagrams, and builds a test matrix from those journeys.

Does ce-dogfood require any external tools to run?

Yes, it requires the agent-browser CLI installed and available on PATH, plus a local dev server command for the project under test. It stops and tells the user to install agent-browser if the binary is missing.

Can ce-dogfood fix the bugs it finds automatically?

It auto-fixes only small, well-understood, low-risk bugs, adding a regression test and committing each fix. Large, architectural, or product-altering issues are escalated to the report's Decisions for a Human section and marked Blocked.

What happens when a test involves OAuth, email, or payments?

External-interaction legs that cannot be driven headlessly are marked Blocked (needs human verify). The skill pauses, asks the user to verify that leg manually, and continues with the rest of the matrix.

Can an interrupted dogfood run be resumed later?

Yes. The report on disk acts as a durable checkpoint with every scenario's status. A later run re-hydrates remaining work from the matrix, keeping passed and fixed scenarios done and surfacing blocked ones for human decision.