exploratory-testing
Skill2 starsChanged 3 months agoMITView on GitHub
Run structured exploratory testing — write a charter, time-box a session, apply heuristics and oracles to discover bugs the scripted suite never looks for, then debrief and file solid bug reports. Use when the user says "test this manually", "find bugs", "explore the feature", "edge cases", "risk-based testing", "session-based", or asks for a charter, a test session, or a bug hunt on a flow that already "works". Framework- and stack-agnostic. Not for writing automated tests — use writing-tests for unit/integration/E2E coverage and regression checks.
13 other projects use a version of this.
Adopted by
Most recently taken up by
- Sn4kyGit/qa-agent-blueprinttheir own version, first seen Nov 16, 2026
- swap-cameroon/kuro-antigravity-configtheir own version, first seen Nov 15, 2026
- jason0971/skillstheir own version, first seen Nov 8, 2026
- Marina-Gajego/qa-skillstheir own version, first seen Nov 1, 2026
- upex-galaxy/agentic-qa-boilerplatetheir own version, first seen Nov 1, 2026
- petrkindlmann/qa-skillstheir own version, first seen Oct 25, 2026
Used with
Projects with it also use ai-system-testing, analytics-tracking-testing and 10 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.
petrkindlmannSkillai-system-testing
Test AI/LLM features that ship in your product. Covers prompt regression testing, response quality evaluation, tool-call validation, hallucination and RAG grounding checks, nondeterministic-output strategies, red-team/safety scans, eval frameworks, and agent-as-target injection (indirect injection via tool output / RAG / scan reports, self-propagating payloads, data exfiltration via an agent) plus a bundled detector for untrusted content. Use when: "test our LLM feature," "prompt regression test," "eval framework," "hallucination test," "RAG grounding," "nondeterministic output," "AI feature testing," "red-team our chatbot," "indirect prompt injection," "agent reading untrusted tool output," "production AI quality." Not for: using AI to generate your own test code — use ai-test-generation. Not for: classifying CI failures with AI — use ai-bug-triage. Not for: EU AI Act / GDPR conformity of an AI feature — use compliance-testing. Not for: canary/flag rollout of an AI feature — use testing-in-production. Related: ai-test-generation, ai-qa-review, api-testing, compliance-testing, security-testing, risk-based-testing, test-data-management.
Also in 3 projects that use this
petrkindlmannSkillanalytics-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
StephanSergioSkilltest-automation
Set up and write automated tests for a vibe-stack app — Vitest + React Testing Library for unit/component, Playwright for E2E, with Firebase and Claude mocked. Use when adding tests, a test harness, or a CI test gate.
Also in 4 projects that use this
conorluddySkillaccessibility-testing
WCAG compliance testing and accessibility quality assurance workflows for iOS apps. Use when validating accessibility labels, testing VoiceOver compatibility, checking contrast ratios, or ensuring WCAG 2.1 compliance. Covers accessibility tree analysis, semantic validation, and automated accessibility testing patterns.
Also in 4 projects that use this
- +8 more from petrkindlmann
Other versions13 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 13
- petrkindlmann/qa-skills★ 274same name
- fishzjp/qa-skills★ 64same name
- upex-galaxy/agentic-qa-boilerplate★ 26same name
- cakehurstryan/QE-Skills★ 15same name
- mattwynne/skills★ 7same name
- Marina-Gajego/qa-skills★ 5same name
- Sn4kyGit/qa-agent-blueprint★ 2same name
- swap-cameroon/kuro-antigravity-config★ 2same name
- jason0971/skills★ 1same name
- mejbaurbahar/software-tester-skills★ 1same name
3 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository19 other files it gives its agents.
- .claude/CLAUDE.mdCLAUDE.md
- api-designSkill
- building-frontend-uiSkill
- debugging-agent-runsSkill
- developing-featuresSkill
- developing-features-tddSkill
- documenting-featuresSkill
- ensuring-cross-platformSkill
- evaluating-ai-outputSkill
- fixing-bugsSkill
- playwright-cliSkill
- reviewing-codeSkill
- running-agent-loopsSkill
- security-reviewSkill
- threat-modelingSkill
- validating-in-realitySkill
- writing-prdSkill
- writing-pull-requestsSkill
- writing-testsSkill
The file
We split this skill into the rules it gives an agent
4 rules. 0 are common, 1 is distinctive, and 3 are very specific to this project.
0
Common
Most of the 11 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 17
If learnings/ is empty except for its README, proceed with the defaults below.
- Distinctive: line 66
Don't burn an exploratory session re-running the scripted happy path — the suite already covers it.
- Very specific: line 67
When exploration finds a recurring or important bug, feed a new automated check back into the suite via writing-tests so it never regresses …
- Very specific: line 68
If a flow can be fully described as fixed inputs and one expected output, it's a check, not an exploration.
“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.
Files
Exploratory Testing
Exploratory testing is simultaneous learning, test design, and execution — you study the product, decide what to try next based on what you just saw, and run the test, all in one tight loop. It is structured, not random. The structure comes from a charter that bounds the search, heuristics that generate ideas, and oracles that tell you when something is wrong.
The goal is to find the bugs nobody wrote a test for: the unknown unknowns. A passing scripted suite proves the known behaviors still hold; it says nothing about the behaviors no one thought to check. That gap is your job.
Read first (always)
List learnings/ and read every file relevant to the current task. Product risk areas, known fragile flows, environment quirks, and "this is always broken under condition X" notes live there and sharpen the charter before you start. If a learning conflicts with this file, the learning wins — mention it to the user.
If learnings/ is empty except for its README, proceed with the defaults below.
The loop
Charter. Frame the session with Elisabeth Hendrickson's formula ("Explore It!"):
Explore (target) with (resources) to discover (information).
Scope it Goldilocks — not too loose, not too tight:
- Too vague: "Test checkout." The session drifts straight to the happy path and finds nothing.
- Too tight: "Click add-to-cart, then click pay, then assert total = $10." That's a script; it kills the agency that makes exploration work.
- Just right: "Explore the payment flow under slow-network conditions to discover state-management failures between discount application and address confirmation." It names a target, a resource that creates pressure, and the class of information you're hunting.
- Time-box a session. 45–90 minutes, uninterrupted. Take reviewable notes as you go (what you tried, what you saw, what surprised you). Notes are the deliverable when no bug is found — they record coverage. See
references/sessions-and-charters.md. - Apply heuristics + oracles. Heuristics generate test ideas (what to try); oracles tell you whether the result is right (how you recognize a problem). Don't improvise from a blank page — run the product through SFDIPOT, vary data deliberately, take a tour. Check every observation against several FEW HICCUPPS oracles. Full catalog in
references/heuristics.md. - File bugs well. A bug nobody can reproduce is noise. Concise summary, numbered steps from a known state, expected vs actual, environment, evidence, and severity and priority noted separately. Template in
references/bug-reports.md. - Debrief with PROOF and branch. After the session, summarize Past (what you did), Results (what you found), Obstacles (what got in the way), Outlook (what's left to explore), Feelings (your read on the risk). Most sessions spawn one or more follow-up charters — file them so the search continues.
Charter discipline
- One charter = one mission. If you find yourself testing three unrelated things, you have three charters; split and queue the others.
- Resources are the lever. "with slow network", "with a screen reader", "with malformed import files", "with two users editing the same record" — the resource is what turns a flat walkthrough into a probe.
- The information clause keeps you honest. If you can't say what class of problem you're hunting, the charter is too vague to start.
Heuristics are the engine
You will not find interesting bugs by clicking around. Pull ideas from structured lists and write down which one you're applying so coverage is visible:
- SFDIPOT — sweep the product's elements: Structure, Function, Data, Interfaces, Platform, Operations, Time.
- Data variation — for every input: boundaries (at / just-below / just-above), counts (zero / one / many), and attacks (empty, null, very long, special chars, Unicode, emoji, negative, leading/trailing spaces, injection).
- Tours — use them as charter generators: follow the money, follow the data (FedEx), the saboteur tour (break things on purpose), the supermodel tour (UI surface only).
Each is detailed, with sources, in references/heuristics.md.
Oracles decide what counts as a bug
An oracle is a principle by which you recognize a problem. Every one is fallible — apply several before you call something a bug. The mnemonic is FEW HICCUPPS (Michael Bolton / DevelopSense): Familiarity, Explainability, World, History, Image, Comparable products, Claims, User expectations, Product (internal consistency), Purpose, Standards/Statutes. When a behavior contradicts one of these, you likely have a bug — describe the inconsistency, not your guess at the cause.
Boundary with automated tests
This is the testing-vs-checking distinction (Bach & Bolton, DevelopSense):
- Checking is confirming a known fact about the product mechanically — that's what automated tests do, and it's what
writing-testscovers. Checks guard behavior you already understand against regression. - Testing is exploring to discover whether the product has problems you didn't anticipate. That's this skill.
Practical rules:
- Don't burn an exploratory session re-running the scripted happy path — the suite already covers it. Go where the suite isn't.
- When exploration finds a recurring or important bug, feed a new automated check back into the suite via
writing-testsso it never regresses again. Exploration discovers; automation defends. - If a flow can be fully described as fixed inputs and one expected output, it's a check, not an exploration. Hand it to
writing-tests.
Capture a learning (final step)
After the task, ask: did I hit a product risk area, an environment trap, a flaky flow, or a charter-scoping convention not in this skill? If yes, append a learnings/YYYY-MM-DD-slug.md (frontmatter description:, body with the rule + Why + How to apply). A learning overrides the defaults here. If no, skip — don't log a summary of the session you just ran.
See also
references/heuristics.md— SFDIPOT, CRUSSPIC STMPL, RCRCRC, FEW HICCUPPS, data-variation patterns, Whittaker's tours, with sources.references/sessions-and-charters.md— SBTM vs Thread-Based TM, charter writing, session-sheet template, PROOF debrief.references/bug-reports.md— bug report template, severity vs priority, triage basics.writing-testsskill — the boundary: automated checks that defend known behavior (hand off regressions discovered here).fixing-bugsskill — once a bug is reproduced, the TDD fix loop (failing test first).reviewing-codeskill — static review of a diff; exploratory testing is the dynamic complement.learnings/— project-specific risk areas, fragile flows, environment quirks.
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
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.

