Candid Chrome QA

Drive a real Chrome session against a running web app, find issues, and write structured findings to disk for downstream triage. Part of the Candid plugin's QA workflow — sits alongside /candid-review (code) and /candid-ship (release).

This is a technique skill. Follow the order. The schema is non-negotiable.

Inputs

When invoked, the user provides (or you confirm):

• goal — surface to test, e.g. "Agent Config / AI Setup, all 16 tabs". Skipped if --goal "<text>" is passed.
• prompt — free-form QA plan (what to exercise, edge cases, hot spots). Skipped if --prompt "<text>" is passed.
• app URL — usually http://localhost:<port>. Verify before you start. Skipped if the user passes --url or chromeQA.defaultUrl is set in .candid/config.json.

CLI flags (full set):