Skip to main content

/testchimp explore — Targeted UX analytics on SmartTests

Use /testchimp explore when ExploreChimp should be the main task: run targeted exploratory analytics on a chosen set of UI SmartTests (Playwright specs that drive a real browser), without running the full PR workflow from /testchimp test.

ExploreChimp is not a separate “mystery shopper” crawl of arbitrary URLs. It reuses the same journeys your SmartTests already execute, then sends structured telemetry to TestChimp so analysis can attribute findings to named screen–state checkpoints.

What “targeted exploratory analytics” means​

  1. Targeted — You (or the agent) pick which spec files and flows matter: a feature area, a failing journey, screens you are shipping in this release, or paths TrueCoverage flagged as high-impact (often combined with /testchimp evolve).

  2. Exploratory — Beyond pass/fail, the pipeline looks for UX-oriented issues: layout and visual regressions, noisy console, suspicious network patterns, accessibility signals (for example via axe), performance / long-task style signals, and similar—along the intervals between your checkpoints.

  3. Analytics — The @testchimp/playwright runtime (with ExploreChimp enabled) records DOM, screenshots, console, optional network (with a tight URL regex you define), and metrics per screen–state interval, correlated by batch id so TestChimp can dedupe and file bugs tied to Atlas vocabulary.

The unit of attribution is the markScreenState fixture: each await markScreenState('Screen', 'State') marks a stable point in the UI; analytics for the prior interval attach to the previous checkpoint, and screenshot + DOM attach to the current one. Without meaningful markers, ExploreChimp has little to hang analysis on—see Screen-State Annotations.

Prerequisites​

  • SmartTests import test / expect from your merged tests/fixtures/index.js (or equivalent), with installTestChimp on that merged test so the markScreenState fixture and reporter wiring apply—see Playwright runtime and Creating Smart Tests.
  • TESTCHIMP_API_KEY in the shell (and any backend URL overrides your repo documents).
  • For ExploreChimp batches: set EXPLORECHIMP_ENABLED, TESTCHIMP_BATCH_INVOCATION_ID (or the project-root id file), and persist network regex / source defaults under plans/knowledge/ai-test-instructions.md → ## ExploreChimp when you use NETWORK analysis.

How this relates to other commands​

CommandRole of ExploreChimp
/testchimp testOptional Phase 5 after Phase 4: Validate, only when the branch plan opts in and the user accepted the extra cost—same runtime flags, ordered after scenario links and markScreenState / atlas hygiene. See Test.
/testchimp evolveOptional targeted runs on TrueCoverage-prioritized UI slices (drop-offs, duration/demand, automation gaps); new tests from the same cycle become valid targets once they pass and have stable markers. See Evolve.
/testchimp explorePrimary exploration: you specify scope, depth, or file list; agent follows the same env and exploratory_runs playbook as the skill.

Product UI and Atlas​

Exploration results surface in TestChimp journeys and Atlas so humans can review bugs and screen–state context. For recipe-style exploration agents in the product, see Exploratory Testing—the /testchimp explore command is the repo + Playwright CLI path aligned with that model.

See also​