# Exploratory Testing > Explore a completed feature before automating it, then read the existing test system and add the minimum sufficient regression coverage in the same context. Use when a new feature is runnable, when automation is about to be written, or when someone asks for exploratory testing. Triggers: "/exploratory-testing", "explore this feature", "test this manually before automating it". - Skill: `haru3613/exploratory-testing` (Agent Skill, multi-file: 3 files) - Install (CLI): `npx skillmds@latest add haru3613/exploratory-testing` - Raw SKILL.md: https://api.skillmd.com/api/skills/haru3613/exploratory-testing/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: haru3613 (https://skillmd.com/u/haru3613) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/haru3613/exploratory-testing --- # exploratory-testing Learn how the feature actually behaves before choosing automation. This skill may edit test code. It never changes expected behaviour to match an implementation. The only permitted product-code edit is the temporary one-line P0 sensitivity injection below, on a local working copy, reverted before anything else. Read repository instructions, the current task, and explicit expected behaviour first. An existing Test Contract is optional context. Missing or stale contract paperwork never prevents exploration or test authoring. Clarify a material ambiguity before encoding it as an automated expectation; continue independent work meanwhile. ## Candidate boundary Use a runnable **non-production** surface: local, preview, simulator, or QA. Record: - full source SHA and local changes, or the actual external candidate identifier; - artifact/build/environment revision and how it maps to that SHA or identifier; - access path, fixtures, permissions, and test data; and - evidence destination. Without trustworthy provenance, observations may guide investigation but cannot become release evidence. Never seed synthetic data into production. ## Pass A — black-box exploration Before reading existing test code, read only the approved behaviour, feature intent, and user-facing surface. Exercise a bounded set of paths chosen by risk: - the P0 happy path; - invalid, empty, boundary, and permission states; - interruption, retry, recovery, or state transitions where applicable; and - interactions with the nearest external boundary. Capture exact steps, observations, console/runtime failures, and durable evidence. If observed behaviour contradicts the intended behaviour, preserve the finding and fix/retest within authorized scope. Use `bug-workflow` when tracking or handoff is needed. Clarify ambiguous expectations before automating them; no revised contract is required. ## Pass B — existing-test audit After exploration: 1. inventory the complete test tree, runners, and CI wiring; 2. deeply read tests for the feature, the same public seam, overlapping user journeys, shared fixtures/helpers, and nearby failure modes; and 3. map every exploration finding to existing coverage before proposing a new case. Large repositories do not require reading every unrelated test file. ## Minimum sufficient automation Choose the cheapest stable seam that catches the failure a user would care about. Extend an existing test or table-driven case before adding another file. Use E2E only when the risk crosses meaningful product boundaries. Any layer—unit, contract, integration, or E2E—is allowed; selection follows risk and seam stability, not role ownership. There is no numeric test limit. Every new case must state: - the distinct behaviour or risk it covers; - the nearest existing test and why its current assertions do not cover the risk; and - the concrete failure that would escape without this case. If that marginal coverage cannot be shown, do not add the case. For critical new coverage, check sensitivity: does the test fail for the wrong behaviour it claims to catch? A meaningful RED observed before a fix already answers that question. Where behaviour already works and sensitivity is uncertain, use a focused temporary fault on an isolated local working copy when safe. Compare the new and nearest existing tests, then revert the fault and rerun the same checks before any other work. Never inject into a shared/deployed surface or commit it. An overly broad fault proves little; aim at the specific forbidden outcome. If existing coverage already catches that exact risk, improve or reuse it rather than adding a redundant test. Record useful sensitivity evidence with the test result. This checks assertion quality; it does not require a contract row, revision, or receipt before a test can count as automation. A missing historical sensitivity record alone is not a release blocker; a demonstrated ineffective assertion is a real coverage gap. Write the minimum sufficient tests in the same context as exploration. Run the changed tests and the nearest relevant regression scope. Run a full suite only when repository policy requires it or the suite is already cheap. Never install a framework without explicit user approval. Choosing the right seam does not make the test at that seam a good one. Follow [behaviour-first tests](tests.md) and [replacing dependencies](mocking.md) for the test's own construction, and [scenario-craft.md](../test-plan/scenario-craft.md) when exploration turns up a risk whose expected or forbidden outcome needs a clearer definition. ## Exploration report Return one concise summary of candidate/environment, paths explored, findings, relevant existing coverage, tests changed, commands/results, and remaining risks. Use the existing PR/issue or chosen location when writes are authorized; separate contract and candidate records are optional. Manual/exploratory observations count for their demonstrated scope. They do not replace explicitly required automation. Reuse unchanged-code evidence only with an equivalence rationale; an earlier run alone does not prove a new artifact or changed environment.