Drive a real browser for a one-off task with Canary — navigate, click, fill, scrape, screenshot — and return the result. Nothing is recorded. Use when the user asks to automate a browser task, scrape a page, fill a form, or check something on a site without needing a report. Trigger phrases — "go to X and get Y", "scrape this page", "automate this browser task", "log in and check", "take a screenshot of".
Open and triage recorded Canary sessions in the local viewer. Use when the user wants to look at, replay, compare, or triage a recorded session — or asks what happened in a run, to open the report, or to see the trace/video/screenshots. Trigger phrases — "open the canary viewer", "show me the last session", "what failed in that run", "review the recording", "open the report".
The Canary sandbox scripting API for browser automation. Use when writing or debugging a Canary script — looking up how to open a page, click, fill, extract text, observe an unknown page with snapshotForAI, evaluate in the page, take a screenshot, persist data between steps, or understand sandbox limits (no imports, timeouts). Trigger phrases — "how do I click in canary", "canary page API", "what's on this page", "explore a page in canary", "snapshotForAI", "saveScreenshot signature", "get text from the page", "why is my canary script timing out", "open a new tab in canary".
Record a verifiable QA session with Canary — explore a flow step by step against one persistent browser, each script a recorded step that captures a Playwright trace, video, network HAR, and console, then render a self-contained report.html. Use when the user wants to verify or QA a flow, produce evidence or a report, or capture a trace/video of a browser run. Trigger phrases — "record a session", "QA this flow", "verify the checkout", "capture a trace", "give me a report of this run".
Turn a code change into a prioritized browser-QA plan with Canary — read the git diff, infer which user-facing workflows it touches, suggest the concrete flows and the checks that must hold, then optionally record those flows as a session with a report.html. Use when the user has changed code and asks what to test, wants to QA a diff, branch, or PR, or wants a focused regression plan before merging. Trigger phrases — "what should I test for this change", "QA my diff", "verify this PR", "I changed X, what flows might break", "regression plan for this branch", "what should I QA before merging".