Skip to content
Agentmart
Submit a tool

dogfood

Formal repro-first browser QA for real web apps.

Skill by MylesMCook

IndexedMIT
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/mylesmcook-dogfood

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Formal repro-first browser QA for real web apps. Use when the user wants to dogfood a site, run QA, bug hunt, find issues, or test an app with handoff-grade evidence such as repro steps, screenshots, videos, and a structured report. Do not use for lighter exploratory passes where `browser-probe` is enough.

Use cases: documents, testing

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/dogfood

What's inside

Skill name
dogfood
Description
Formal repro-first browser QA for real web apps. Use when the user wants to dogfood a site, run QA, bug hunt, find issues, or test an app with handoff-grade evidence such as repro steps, screenshots, videos, and a structured report. Do not use for lighter exploratory passes where `browser-probe` is enough.
License
MIT
Instructions
49 lines in SKILL.md · view on GitHub

Files (3)

  • SKILL.md
  • issue-taxonomy.md · references/
  • dogfood-report-template.md · templates/

SKILL.md (first lines)

# Dogfood

Use this skill when the job is not just to explore, but to hand back issues with proof.

## Defaults

- Require only the target URL unless auth is mentioned.
- Use `agent-browser` directly, never `npx agent-browser`.
- Default scope is the full app.
- Default output directory is `./dogfood-output/`.
- Default session name is a slugified target name.

## Workflow

1. Initialize the output directory, screenshots directory, videos directory, and report file from `templates/dogfood-report-template.md`.
2. Authenticate only if needed. If the flow needs OTP or email codes, stop and ask the user for that code.
3. Orient on the app with an annotated screenshot and a fresh interactive snapshot.
4. Explore the app systematically using `references/issue-taxonomy.md` to decide what to check and how severe a finding is.
5. Document each issue immediately before moving on. Do not batch findings for later.
6. Wrap up by reconciling severity counts, closing the session, and summarizing the most important issues.

## Evidence Rules

- Interactive or behavioral issues need a repro video plus step-by-step screenshots.
- Static issues visible on load need one annotated screenshot. Set repro video to `N/A`.
- Verify reproducibility before collecting evidence. Retry once before calling it a real issue.
- Use `type` during recorded repros when human pacing matters. Use `fill` outside recording when speed matters.
- Check `console` and `errors` during the run so console-visible failures do not get missed.

## Output Contract

- Produce one structured report using `templates/dogfood-report-template.md`.
- Append findings as they are discovered so interrupted runs still leave usable output.
- Aim for a small set of well-documented issues rather than a long vague list.
- End by telling the user the report path, total issue count, severity breakdown, and the most critical items.

## References

- `references/issue-taxonomy.md` - severity calibration and what to look for
- `templates/dogfood-report-template.md` - report scaffold

## Guardrails

- Keep one live `agent-browser` session for the run unless recovery forces a restart.
- Never read the target app's source code while dogfooding it.
- Do not turn this into generic browser automation. This skill is for issue-finding with handoff-grade evidence.
- Route lighter exploratory requests to `browser-probe`.

From MylesMCook/mcook-skills, MIT.

More from MylesMCook/mcook-skills

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Apr 22, 2026 (6 months ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from MylesMCook

Cite this page

dogfood skill. Agent Mart. https://agentmart.dev/skills/mylesmcook-dogfood (last updated 2026-10-24)