agentleFS
Sn4kyGitSn4kyGit/qa-agent-blueprint/.claude/skills/exploratory-testing/SKILL.md

exploratory-testing

Skill2 starsChanged 7 days agoApache-2.0View on GitHub

Run a chartered, SBTM-style exploratory testing session via Playwright MCP — for new features, unclear behavior, or bug reproduction without a repeatable spec. Use for "/exploratory-testing", "erkunde das neue Feature", "explorativ testen", "reproduzier den Bug". Not for understood, repeatable flows — use test-automation.

13 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use analytics-tracking-testing, compliance-testing and 4 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • petrkindlmannSkill

    analytics-tracking-testing

    Validate that analytics and marketing tracking fire CORRECTLY: GA4/GTM dataLayer events, Meta/TikTok/LinkedIn pixels, and ad-tech tags. Covers building a tracking plan as the contract, intercepting collect-endpoint beacons and dataLayer.push in Playwright, asserting event name + params + values + timing + de-duplication, Consent Mode v2 gating, CI regression gating, and news-media events (article-view, scroll-depth, paywall). Use when: "test analytics tracking," "GA4 event test," "verify the pixel fires," "dataLayer test," "tracking plan," "Meta Pixel dedup," "scroll-depth tracking test," "gate tracking in CI." Not for: whether tracking is ALLOWED to fire under consent law (GDPR/CMP) — that is compliance-testing; this skill checks the data is CORRECT. SEO meta tags / structured data — out of scope. Related: compliance-testing, playwright-automation, api-testing, qa-project-context.

    Also in 3 projects that use this

  • petrkindlmannSkill

    compliance-testing

    Test for regulatory compliance: GDPR/CMP consent verification, Google Consent Mode v2, Global Privacy Control (GPC), CCPA/US state opt-out, EU AI Act Article 50 transparency, Better Ads Standards, and cookie-inventory auditing. Covers automated consent-flow testing, policy-based script blocking before consent, and cookie drift detection. Use when: "GDPR test," "compliance," "CMP test," "cookie consent," "consent mode," "CCPA," "GPC," "AI Act," "Better Ads," "privacy banner." Not for: WCAG/axe-core test authoring — use accessibility-testing. Not for: OWASP/vuln scanning — use security-testing. Not for: evaluating your LLM feature's quality or safety — use ai-system-testing. Related: accessibility-testing, security-testing, ai-system-testing, ci-cd-integration.

    Also in 3 projects that use this

  • jarosser06Skill

    test-strategy

    Plans pytest test strategies for Drift's 90%+ coverage requirement using coverage analysis, edge case identification, and test suite design. Use when determining what tests to write, analyzing coverage gaps, or planning test strategies.

    Also in 3 projects that use this

  • +3 more from petrkindlmann

The file

We split this skill into the rules it gives an agent

8 rules. 0 are common, 3 are distinctive, and 5 are very specific to this project.

  • 0

    Common

    Most of the 11 edited versions we read keep these.

  • 3

    Distinctive

    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. Very specific:

    When a run-scoped context-capsule.json path is supplied, use it as the effective context for this session: exact ACs, routing, …

    line 13
  2. Very specific:

    Run the preflight first (same gate as test-ticket): the blueprint's preflight.sh <KEY> must pass before any browser/JIRA call.

    line 23
  3. Distinctive:

    Write the charter: before starting, in the SBTM pattern:

    line 27
  4. Very specific:

    Prefer API/CLI and write readiness.json.

    line 52
  5. Distinctive:

    Only then may the stage-liveness navigation and this session begin.

    line 56
  6. Very specific:

    If the page registers WebMCP tools, tag them as NOTE/RISK in the session log and name them in the debrief as candidates for deeper coverage …

    line 103
  7. Distinctive:

    Exploration-to-automation pipeline: every reproducible bug found here should graduate into an automated regression test, so the *next* …

    line 113
  8. Very specific:

    If a spec already exists for the exact flow, or the check is meant to run repeatedly (nightly, every retest), use test-automation instead — …

    line 127

“Common” is measured: kept by at least 6 in 10 of the 11 edited versions of this file we could read. “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

/exploratory-testing — chartered session via Playwright MCP

Session-Based Test Management (SBTM) applied to our ticket-driven flow: a written charter, a time-boxed session delegated to test-executor, heuristic-guided exploration, and a debrief — not "click around and see what happens."

When a run-scoped context-capsule.json path is supplied, use it as the effective context for this session: exact ACs, routing, stage/test-user references, confirmed Areas and relevant Learnings. Do not fully read PROJECT/PRODUCT, route again or repeat the mode/suite/Learnings merge. Pass the capsule path to test-executor. Targeted new symptom/flow Learning reads remain allowed. A stale capsule returns STALE_CONTEXT to the orchestrator; never regenerate it here. Without a capsule, the standalone fallback below reads the effective project context as before.

Steps

  1. Run the preflight first (same gate as test-ticket): the blueprint's preflight.sh <KEY> must pass before any browser/JIRA call. Do not use a stage navigation as an early liveness probe. JIRA access browser may use Playwright for JIRA intake only.
  2. Write the charter before starting, in the SBTM pattern:

    Explore [target]
      with [resources]
      to discover [information]

    E.g. "Explore the checkout flow's coupon handling, with the qa-standard test user and three coupon states (valid/expired/exhausted), to discover requirement gaps against SHOP-1234's acceptance criteria." Not "test the app." With a capsule, use its selected knowledge. Without one, check memory/learnings/*.md and projects/<KEY>/learnings.md for known risk areas or past bug clusters in this area before finalizing the target. When a flow map exists (projects/<KEY>/[products/<PRODUCT>/]flow-map.md, or the bridge's ## Flow map): confirmed Areas whose last confirmed is oldest are charter candidates — an unconfirmed flow is a hypothesis, not a truth — and the map names the target area precisely. Unconfirmed candidates supply no binding anchors.

  3. Confirm project context is loaded: test users, stage URL, environment gotchas. Take them from the supplied capsule; only a standalone run without a capsule reads PROJECT.md plus the routed product's PRODUCT.md. Name the product in the charter when one is routed. 2b. For a ticket-driven session, finish intake/routing and the readiness contract (docs/pipeline/01a-readiness.md) before the first stage navigation: allowed status, no blockers, required linked merge complete, expected build deployed when verification is required (otherwise explicitly configured and reported unverified). Prefer API/CLI and write readiness.json; missing merge/required build means BLOCKED or a deploy handoff, not exploratory clicking. An area-only charter needs an explicitly authorized target and, under required verification, an expected build; unavailable verification stays documented as unverified. Only then may the stage-liveness navigation and this session begin.
  4. Time-box it: 45–90 minutes of actual exploration. Bug-finding effectiveness drops sharply past 90 minutes — split a bigger area into multiple chartered sessions rather than one long unfocused one.
  5. Delegate the session to subagent test-executor with the charter and, when available, the sealed capsule path rather than pasted context. Use test-executor-deep when reliable capsule intake marks high risk or this branch is escalated; otherwise keep the standard agent. Missing or unknown risk never selects a cheap or deep lane, and this mode/escalation decision remains separate from model selection. The selected agent drives the configured playwright MCP browser (headless/headful and isolation per the user's .playwright-mcp.config.json) per docs/rules/testing/playwright-mcp.md and produces evidence per .claude/rules/core/evidence-policy.md. When the session is a recon run (pipeline phase 2, the /recon MCP path), the executor additionally writes recon.json into the run dir at its end — pages, controls, form fields, console errors, failed requests in the probe format (docs/pipeline/02-recon.md, "source": "mcp"), so the area has its phase-2 artifact like a probe run. This flow has no summary step of its own: whoever writes the run's execution_summary.json lists exploratory-testing (and any other skill the session actually used, never merely planned ones) in skills_used. Instruct it to explore through the HICCUPS lenses, not just the happy path:

    LetterOracleAsk
    HHistoryDid this work before? Has behavior changed subtly?
    IImageDoes it look/feel as polished as the rest of the product?
    CComparableHow do similar products/flows handle this?
    CClaimsDoes it match the ticket's acceptance criteria / spec / tooltip text?
    UUser expectationsWould a real user find this confusing?
    PProductConsistent with the rest of this product (e.g. other error messages)?
    SStandardsWCAG, RFC, GDPR — whatever standard applies here?

    For concrete boundary/state-transition/error-handling/"what if" test-idea banks, see references/heuristics-and-automation.md. For the session flow (orient → explore → wrap-up timing), a session-log template with BUG/QUESTION/IDEA/RISK/NOTE tags, and worked charter examples, see references/session-templates.md.

    Mandatory in the orient instructions: the WebMCP capability probe. Once the page under test is loaded, exactly one browser_evaluate call:

    typeof document.modelContext !== 'undefined'
      ? await document.modelContext.getTools()
      : null

    If the page registers WebMCP tools, tag them as NOTE/RISK in the session log and name them in the debrief as candidates for deeper coverage (schema validity, readOnlyHint honesty, exposedTo/permissions audit). Never use these tools to drive the session — the UI stays the test target and is driven via Playwright; WebMCP tools are extra surface, not a shortcut past it. An orient phase without this probe is incomplete.

  6. Debrief, always — even a solo session. Tag every observation (BUG/QUESTION/IDEA/RISK/NOTE) with a timestamp as it happens, not from memory afterward. Mark each charter area covered, partial, or unexplored — an honest "partial" beats a false "100% covered."
  7. Exploration-to-automation pipeline: every reproducible bug found here should graduate into an automated regression test, so the next session explores new ground instead of re-checking this one. If the flow itself (not just the bug) is worth repeating going forward, hand it to test-automation — see the worked example in references/heuristics-and-automation.md.
  8. On a FAIL, the orchestrator running this skill invokes failure-analyst with the failure packet and, when available, the capsule path before any bug handoff; test-executor never delegates to the analyst. A confirmed product-defect/undetermined finding of severity P0–P2 then goes to bug-report; other classes follow their normal BLOCKED/NEEDS_REVIEW/reliability path. Anything unexpected about the environment or tooling itself → capture via retro-learn.

When NOT to use this

If a spec already exists for the exact flow, or the check is meant to run repeatedly (nightly, every retest), use test-automation instead — exploratory sessions are for understanding and bug-hunting, not for repeatable verification. For risk scoring of which areas deserve a chartered session, see risk-based-testing. For turning a vague, already-reported bug into a minimal, deterministic repro, see bug-reproduction instead of a fresh exploratory charter.

Anti-patterns

  • Unchartered exploration — "poke around and see" produces nothing repeatable or debriefable. Always write the one-sentence charter first.
  • No notes during the session — observations not recorded in the moment are lost by the time the session ends. Use the session-log template, not memory.
  • Happy-path-only exploration — that duplicates what the automated suite already covers. Use the HICCUPS lenses to push into uncomfortable territory.
  • Bugs found but never converted — finding the same bug manually every release because step 6 never happened.

Charter model, HICCUPS/FEW HICCUPS heuristics, and the two reference files in references/ are adapted from the exploratory-testing skill in petrkindlmann/qa-skills (MIT license) — see .claude/skills/THIRD_PARTY_NOTICES.md.

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.