agentleFS
timeyourtimeyour/agentskills-audit-collection/.claude/skills/physical-flow-test/SKILL.md

physical-flow-test

Skill1 starsChanged 6 months agoMITView on GitHub

Generate executable real-browser flow tests using Python Playwright from business flows. Use when cognitive audit, flow-test, audit, or accept-five needs physical browser proof for CTA, form, auth, upload, checkout, dashboard, production, staging, or local workflows.

  • Reads credentials

The file

We split this skill into the rules it gives an agent

11 rules. About 0 are common, 6 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill to turn business flows into real executable browser tests.

    line 11
  2. Very specific:

    Use /physical-flow-test when the user needs to:

    line 21
  3. Distinctive:

    Verify critical user paths with real browser clicks.

    line 23
  4. Distinctive:

    Check CTA, forms, login, signup, dashboard actions, uploads, exports, checkout, or booking flows.

    line 28
  5. Very specific:

    Prefer /harness first. If /harness is unavailable, perform a minimal harness decomposition inside this skill before generating code.

    line 31
  6. Distinctive:

    Never claim a workflow works without real execution evidence.

    line 35
  7. Very specific:

    Use python -m pytest -q tests/, not the Node/TypeScript Playwright runner, unless the user explicitly asks for TypeScript.

    line 38
  8. Distinctive:

    Every generated test must include trace capture, screenshot on failure, HAR/network capture, video when supported, console log capture, …

    line 39
  9. Distinctive:

    Use the audit permission model before generating tests for production, authenticated, payment, deletion, upload, admin, or private-data …

    line 41
  10. Very specific:

    Never hardcode secrets, passwords, tokens, or cookies.

    line 43
  11. Very specific:

    Use S0-S4 severity and preserve the shared output shape.

    line 45

≈ “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

Physical Flow Test

Use this skill to turn business flows into real executable browser tests.

business flow -> harness decomposition -> Python Playwright package -> real execution -> artifacts -> regression check -> lessons

This skill must not guess whether a workflow works. It produces tests that a user, CI job, or real browser environment must execute.

When To Use

Use /physical-flow-test when the user needs to:

  1. Verify critical user paths with real browser clicks.
  2. Upgrade /flow-test from static inspection to physical execution.
  3. Generate repeatable regression tests for important workflows.
  4. Validate staging, production, or local environments with real network behavior.
  5. Collect physical evidence for /audit or /accept-five.
  6. Check CTA, forms, login, signup, dashboard actions, uploads, exports, checkout, or booking flows.
  7. Convert business-level acceptance criteria into runnable browser tests.

Prefer /harness first. If /harness is unavailable, perform a minimal harness decomposition inside this skill before generating code.

Core Rules

  1. Never claim a workflow works without real execution evidence.
  2. Decompose the business goal into minimal executable units before writing tests.
  3. Generate Python tests using pytest and playwright.sync_api.
  4. Use python -m pytest -q tests/, not the Node/TypeScript Playwright runner, unless the user explicitly asks for TypeScript.
  5. Every generated test must include trace capture, screenshot on failure, HAR/network capture, video when supported, console log capture, clear assertions, and safe failure handling.
  6. Output generated packages under ./artifacts/physical-tests/{timestamp}-{flow-name}/.
  7. Use the audit permission model before generating tests for production, authenticated, payment, deletion, upload, admin, or private-data flows.
  8. Mark payment, deletion, irreversible submission, production email/SMS sending, and real purchase flows as SKIPPED-SAFE unless explicitly allowed.
  9. Never hardcode secrets, passwords, tokens, or cookies. Use environment variables and .env.example placeholders.
  10. Require artifact redaction before sharing HAR, trace, screenshots, video, or console logs.
  11. Use S0-S4 severity and preserve the shared output shape.

Workflow

  1. Intake and scope: identify flow name, URL or local command, environment, browser target, auth needs, test account, high-risk actions, expected results, and required artifacts.
  2. Surface and permission check: use any available web surface map, apply the permission model, and mark unsafe units SKIPPED-SAFE.
  3. Harness decomposition: map business goal -> user intent -> page/route -> action -> locator -> expected result -> checkpoint -> failure signal.
  4. Generate test package: create README.md, requirements.txt, .env.example, run-tests.sh, tests/test_{flow_name}.py, and artifact directories.
  5. Execution instruction: provide exact commands for installing Python dependencies, installing Chromium, and running pytest.
  6. Artifact review: inspect returned trace, screenshots, HAR, console logs, video, and result JSON against expected behavior.
  7. Regression and lessons: convert findings into regression checks, locator rules, auth/session rules, network dependency rules, timeout rules, and safe-skip rules.

References

  • references/python-playwright-template.md
  • references/artifact-schema.md
  • references/safe-execution-policy.md
  • references/locator-policy.md
  • references/regression-lessons-ledger.md
  • ../audit/references/permission-model.md
  • ../audit/references/web-surface-discovery.md when a surface map is needed before test generation

Output Format

Physical Flow Test Summary
Target Flow:
Environment:
Harness Used:
Execution Mode:
Risk Level:

Scope
Evidence
Findings
Severity
Reproduction
Fix Suggestion
Regression Check
Lessons

Anti-Patterns

  • Claiming a workflow works without physical execution.
  • Mentally simulating browser clicks.
  • Generating tests without requiring execution.
  • Testing only happy paths.
  • Missing trace, HAR, video, console log, or failure screenshot capture.
  • Hardcoding credentials.
  • Sharing raw HAR files without redaction.
  • Running destructive production actions by default.
  • Using brittle selectors when accessible locators exist.
  • Treating docs, marketing copy, source code, or screenshots as proof of working behavior.

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.