← ClaudeAtlas

exploratory-charterlisted

Generate session-based exploratory testing charters and run the session properly — a ranked set of missions with heuristics and oracles, plus a session sheet that records what was actually learned. Use when scripted cases are not enough, when testing something new or poorly specified, when asked to "have a look at" a feature, or when a bug is suspected but not yet found.
rcptrkr/qa-skills · ★ 1 · Testing & QA · score 63
Install: claude install-skill rcptrkr/qa-skills
# Exploratory Charter Exploratory testing is not unstructured clicking, and it is not a substitute for having thought. It is **simultaneous learning, test design, and execution**, kept accountable by two artifacts: a *charter* that states the mission before you start, and a *session sheet* that records what you found after. Scripted tests confirm what someone already imagined. Exploration finds what nobody imagined — which is where the defects that reach production actually come from. ## Charter format Three parts, one sentence: > **Explore** _target_ **with** _resources_ **to discover** _information_. The third part is the one people get wrong. "To discover bugs" is not information — it is a wish. Name the *question* the session answers. Good: - Explore **the bulk CSV import** with **files that are malformed in one way each** to discover **how partial failures are reported and whether partial data is committed**. - Explore **session expiry** with **two browser tabs and a clock skewed forward** to discover **whether an expired session in one tab can still act through the other**. - Explore **the order search API** with **Turkish locale data (İ/ı, ç, ş) and 10k-row result sets** to discover **where casing and paging assumptions break**. Bad: - Explore the checkout page to find bugs. *(no question, no resources, unbounded)* - Test the new report screen. *(not a charter — a task)* "Bugs" is the obvious failure. The common one is subtler: naming a *category* of defect an