Skip to main content
affpapa.org
RU EN
skill Claude Code CX Codex Cursor GitHub Copilot

dogfood · callstack — Claude Code skill: what it does, how to install

Coordinates exploratory mobile app testing on iOS and Android via agent-device. Discovers UI bugs and workflow issues in runtime.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓ 
npx skills add callstack/agent-device --skill dogfood -a claude-code
Open on GitHub

What dogfood does

The skill manages exploratory QA cycles on iOS and Android builds. Reported issues come strictly from live runtime execution rather than static source code analysis. Its workflow follows a clear loop: launch the target app, inspect the view tree via snapshot -i, take a screenshot, step through user flows, collect evidence per issue, and close the app.

Tool execution is strictly restricted to agent-device and npx agent-device in Bash. The skill requires agent-device >= 0.14.0. If the binary is missing from PATH or outdated, the assistant halts immediately instead of silently upgrading. It will ask for a trusted binary path or an exact-version npm command approval because running autonomous @latest npm installs is forbidden.

Report formatting and evidence commands come directly from agent-device help dogfood. Specifying a target app is required, while the platform can be inferred or clarified during the run.

What it helps with

  • explore UI flows and detect bugs on iOS and Android
  • capture screenshots and accessibility snapshots as bug evidence
  • run QA based on runtime app behavior without reading source files

Example prompts for the agent

  • Run dogfood on the Android app to test checkout flows and report UX bugs
  • Use dogfood on iOS to do exploratory QA and capture evidence

How to install dogfood

1Skills CLI (Vercel)

npx skills add callstack/agent-device --skill dogfood -a claude-code

2Manually: put the files in place

~/.claude/skills/dogfood/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/dogfood/

From the same set2

The whole set →
agent-device callstack This skill lets an agent navigate apps on Apple platforms and Android, including TV devices. Use it to interact with screens, enter text, and inspect app state. Claude Code CX Codex Cursor +1 18,700installs

Other versions of dogfood

dogfood Teknium (teknium1), Hermes Agent A skill for exploratory QA of a web app: it walks the pages, captures console errors and screenshots, then writes a bug report. Claude Code CX Codex Cursor +1 6,400installs

Similar

llm-evaluation Seth Hobson A skill for evaluating LLM applications: automated metrics, human feedback, LLM-as-judge and benchmarking. Useful when comparing models or catching regressions before deploy. Claude Code CX Codex Cursor 12,755installs protocol-reverse-engineering Dávid Balatoni This skill helps inspect network traffic, map protocol structures, and write protocol documentation. Use it for proprietary protocol analysis or network debugging. Claude Code CX Codex Cursor 12,701installs cmux-testing manaflow-ai Skill for choosing verification scopes, adding tests, and validating Swift test targets in the cmux project. Decides what local and CI evidence a change needs. Claude Code CX Codex Cursor +1 4,900installs wcag-audit-patterns Seth Hobson Audits web interfaces against WCAG 2.2 with automated checks and manual verification. Use it to find violations and build accessible patterns. Claude Code CX Codex Cursor 12,486installs cmux-debugging manaflow-ai Skill for adding debug logging, diagnosing UI and runtime issues in cmux. Use when working with terminal rendering, tab lists, drag-and-drop, or testing across macOS versions. Claude Code CX Codex Cursor +1 4,700installs dfam-check earthtojake Measures mesh files against Design for Additive Manufacturing rules to evaluate printability before slicing or running CAD regeneration. Claude Code CX Codex Cursor +1 4,100installs improve-codebase-architecture Matt Pocock This skill scans a codebase for architectural friction, then builds a visual HTML report with refactoring candidates. Use it before a deeper architecture review. Claude Code 1,163,035installs design-taste-frontend Leonxlnx This skill directs the design of public pages and work screens, then checks the result in a browser through screenshots. Use it when an interface feels generic or unfinished. Claude Code CX Codex Cursor +1 673,100installs

Questions about dogfood

How to install dogfood in Claude Code?

Install it by running npx skills add callstack/agent-device --skill dogfood -a claude-code.

What does the dogfood skill do?

It systematically explores mobile apps on iOS and Android, taking snapshots and screenshots to identify UI glitches and workflow bugs.

What are the agent-device requirements?

It requires agent-device >= 0.14.0. If missing or outdated, execution stops, requiring a trusted PATH binary or an approved exact-version npm command without autonomous latest upgrades.

Source text

SKILL.md · MIT 
---
name: dogfood
description: Systematically explore and test a mobile app on iOS/Android with agent-device to find bugs, UX issues, and other problems. Use when asked to dogfood, QA, exploratory test, find issues, bug hunt, or test this app on mobile.
allowed-tools: Bash(agent-device:*), Bash(npx agent-device:*)
---

# Dogfood

Router for exploratory QA. Private setup before using this skill:

```bash
agent-device --version
```

If that fails, stop and tell the user to expose a trusted `agent-device` binary on PATH or approve an exact-version npm command. This skill intentionally keeps allowed tools restricted to `agent-device` and `npx agent-device`.

Require `agent-device >= 0.14.0`; older CLIs lack these help topics. If older, stop and tell the user to upgrade the trusted install or approve an exact-version npm command. Do not run `npm install -g agent-device@latest` or `npx -y agent-device@latest` autonomously, and do not include version/upgrade commands in final plans.

Read current CLI guidance:

```bash
agent-device help dogfood
```

Loop: open app -> snapshot -i + screenshot -> explore flows -> capture evidence per issue -> close.

Target app is required; infer platform or ask. Findings must come from runtime behavior, not source reads. Let `help dogfood` provide exact report shape, evidence commands, and current workflow guidance.

© callstack · MIT · callstack/agent-device

See also

Project sponsors

Companies that keep this analytics open
Gold sponsor
Nutra.Club

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor