agentleFS
ulpi-ioulpi-io/browse/.agents/skills/browse-qa/SKILL.md

browse-qa

Skill15 starsChanged 8 months agoApache-2.0View on GitHub

QA a web or app feature from a ticket, URL, acceptance criteria, or plain-language description, then generate reusable browse flows for regression coverage. Uses the `browse` skill for the actual browser or simulator interaction and keeps the QA loop focused on scenarios, evidence, and report quality.

The file

Show its rules and text

We split this skill into the rules it gives an agent

13 rules. About 0 are common, 7 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use the browse skill for actual browser or simulator interaction.

    line 32
  2. Distinctive:

    If the target or platform is ambiguous, ask before testing.

    line 62
  3. Distinctive:

    Keep the plan small enough to execute in one session unless the user asked for a broad QA sweep.

    line 75
  4. Very specific:

    Use Skill to invoke browse for:

    line 81
  5. Very specific:

    keep the same browse session for one logical user journey unless isolation is intentional

    line 91
  6. Distinctive:

    use headed or handoff mode only for real blockers like CAPTCHA, MFA, or OAuth walls

    line 92
  7. Distinctive:

    When the scenario is stable and worth rerunning:

    line 99
  8. Distinctive:

    Do not save noisy or half-manual recordings as regression artifacts.

    line 105
  9. Very specific:

    Do not duplicate the full browse command manual inline.

    line 128
  10. Distinctive:

    Do not guess a platform when the request is ambiguous.

    line 129
  11. Distinctive:

    Do not save low-signal flows just to say a regression asset was created.

    line 130
  12. Very specific:

    Do not add disable-model-invocation.

    line 131
  13. Very specific:

    Do not add context: fork; results are usually needed in the current execution flow.

    line 132

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md
<EXTREMELY-IMPORTANT> This skill owns the QA plan and report, not the low-level browser command encyclopedia.

Non-negotiable rules:

  1. Turn the incoming spec into explicit scenarios before testing.
  2. Confirm ambiguous targets or platforms before execution.
  3. Use the browse skill for actual browser or simulator interaction.
  4. Capture evidence for each failed or important scenario.
  5. Save rerunnable flows only when they are cleanly scoped and valuable for regression.</EXTREMELY-IMPORTANT>

browse-qa

Inputs

  • $request: Ticket, URL, app target, acceptance criteria, or plain-language QA request

Goal

Produce a credible QA run that:

  • turns a vague request into explicit scenarios
  • runs those scenarios against the real target
  • captures evidence for passes and failures
  • saves rerunnable browse flows when appropriate
  • returns a clean QA report the user can act on

Step 0: Resolve the target and spec

Work out:

  • what feature is being tested
  • the platform: web, iOS simulator, Android emulator, or macOS app
  • the target URL, build artifact, or app identifier
  • the acceptance criteria or expected outcomes

If the target or platform is ambiguous, ask before testing.

Success criteria: The QA target and acceptance criteria are explicit.

Step 1: Build the scenario list

Break the request into:

  • happy path
  • validation or error path
  • edge cases
  • exploratory probes worth trying after the stated acceptance criteria

Keep the plan small enough to execute in one session unless the user asked for a broad QA sweep.

Success criteria: There is a concrete scenario list with expected outcomes.

Step 2: Execute with the browse skill

Use Skill to invoke browse for:

  • navigation
  • page stabilization
  • interactive element discovery
  • clicks, fills, and submissions
  • screenshots, console logs, and network inspection

Rules:

  • keep the same browse session for one logical user journey unless isolation is intentional
  • use headed or handoff mode only for real blockers like CAPTCHA, MFA, or OAuth walls
  • save screenshots and flow artifacts under the browse session directories

Success criteria: Each scenario is executed against the real target with enough evidence to judge pass or fail.

Step 3: Record reusable regression flows

When the scenario is stable and worth rerunning:

  • record the flow
  • split unrelated user journeys into separate flow files
  • give each saved flow a descriptive name

Do not save noisy or half-manual recordings as regression artifacts.

Success criteria: Saved flows are independently rerunnable and map to clear scenarios.

Step 4: Report results

For each scenario, record:

  • expected result
  • actual result
  • pass or fail
  • evidence such as screenshot path, console error, or network symptom

Also report:

  • exploratory findings
  • regression flow files created
  • blockers that prevented full verification

Success criteria: Another engineer can understand the QA outcome without replaying the whole session first.

Guardrails

  • Do not duplicate the full browse command manual inline; use the browse skill.
  • Do not guess a platform when the request is ambiguous.
  • Do not save low-signal flows just to say a regression asset was created.
  • Do not add disable-model-invocation; QA should stay available when the user asks for it.
  • Do not add context: fork; results are usually needed in the current execution flow.

Output Contract

Report:

  1. target tested
  2. scenarios executed
  3. pass/fail result for each scenario
  4. evidence captured
  5. regression flow files created
  6. blockers or uncovered gaps

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.