Exploratory Testing Skill for Autonomous Agent Explorations

Where this fits

A developer hands an agent a half-written requirement, a staging URL, and a login. No docs, no test cases, no clear behavior model. The agent needs to understand the system before it can design tests or write automation. This skill enables a charter-driven exploration session that maps the system and surfaces risks, producing an exploration note that feeds requirement analysis or stands alone as a deliverable.

Why agents benefit

• Charter focus: the goal → explore → record structure keeps an agent from wandering across unrelated areas.
• Structured output: the exploration note schema turns observations into reusable artifacts — system understanding, risk list, test ideas.
• Evidence discipline: each risk carries an evidence tag and confidence level, so downstream agents can judge how much to trust it.
• Explicit boundaries: the "when NOT to use" rules prevent an agent from using this skill for automation setup or executing existing cases.
• Black-box readiness: works with only an entry point and credentials, which is exactly the environment an agent often gets.

Practical scenarios

Onboarding to an unfamiliar system

An agent is asked to add tests for a legacy admin portal. The frontend is undocumented and the API contracts are stale. Instead of guessing, the agent starts an exploratory session with a charter like "map role-based permissions across the user management module" and produces a system understanding note that later guides test design.

Filling gaps in incomplete requirements

The ticket says "checkout total should be correct" but says nothing about discounts, taxes, or currency. The agent explores the checkout flow, records observed state transitions, and lists risks such as missing tax edge cases. That note becomes a concrete input for requirement analysis.

Verifying suspected fragile areas

A developer suspects the refund flow breaks after a recent payment provider change. The agent runs a time-boxed exploration focused only on refund paths, collecting evidence for each risk instead of blindly executing existing happy-path tests.

Add it to your agent workflow

Install it with:

npx skills add fishzjp/qa-skills --skill exploratory-testing

Then invoke it with an environment, account, and charter:

Input:
  environment: staging
  account: demo@example.com
  charter: Map the order creation flow and identify payment-related risks
Output:
  exploration-note_order-creation.md

You can also see the OpenAgentSkill listing at https://www.openagentskill.com/skills/fishzjp-exploratory-testing for updated usage notes.

Compare before adopting

Look at how this skill enforces evidence tags and a clear output schema versus skills that just say "explore and report." Check the repository for recent commits and maintenance freshness — 12 stars is small, so verify the author is still active. Compare it with automated-e2e-testing if your real goal is selector extraction, or api-testing if you already have scripted cases. Most importantly, evaluate whether the exploration note format fits your existing requirement-analysis workflow.

Why it is worth tracking

This skill is a focused pattern for a real agent problem: acting under incomplete information. Its versioned README, explicit charter structure, and evidence-based risk list make it easy to evaluate and adopt. The community momentum is modest, but the design quality is high. Watch the repo for further development and reassess it as the qa-skills collection matures.