Feat/qa agent - #92
Conversation
Sorry, something went wrong.
|
apple.com demo thebiguglywebsite.com demo |
Sorry, something went wrong.
| category: "compliance", | ||
| complianceType: "accessibility", | ||
| standard: issue.standard, | ||
| description: issue.description, |
There was a problem hiding this comment.
Missing AI response validation crashes report generation
Medium Severity
When constructing issues from AI responses, description: issue.description is assigned directly without a fallback (unlike severity which uses || "info"). Since parseAIResponse returns any[], the AI could return objects missing the description field. Later, escapeHtml(issue.description) is called unconditionally in renderIssue, and calling text.replace() on undefined throws a TypeError. This crashes report generation even when all analysis checks completed successfully.
Additional Locations (1)
Sorry, something went wrong.
| version: 1.57.0 | ||
| zod: | ||
| specifier: ^4.2.0 | ||
| version: 4.3.5 |
There was a problem hiding this comment.
Lock file mismatches package.json dependency versions
Medium Severity
The pnpm-lock.yaml specifiers don't match package.json dependencies. Key mismatches: @anthropic-ai/sdk is ^0.52.0 in package.json but ^0.30.1 in pnpm-lock.yaml; zod is ^3.23.8 in package.json but ^4.2.0 in pnpm-lock.yaml (major version difference with breaking changes); luxon is in package.json but missing from pnpm-lock.yaml; playwright-core appears in pnpm-lock.yaml but not in package.json. Users installing with pnpm will get unexpected versions or installation failures.
Additional Locations (1)
Sorry, something went wrong.
… any llm for vision
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.
Sorry, something went wrong.
|
|
||
| async stop(): Promise<SessionInfo> { | ||
| const info = this.info; | ||
|
|
There was a problem hiding this comment.
Session stop() throws before checking if session exists
Medium Severity
The stop() method calls this.info on line 85 before checking if (this._sessionId) on line 87. The info getter accesses this.sessionId, which throws an error if _sessionId is null. This means calling stop() twice or calling it without first calling start() will throw an unhelpful "Session not started" error instead of being handled gracefully by the existing null check. The control flow is inverted - the safety check happens after the code that needs protection.
Sorry, something went wrong.
Reviewers
At least 1 approving review is required to merge this pull request.
Note
Medium Risk
Mostly additive template code, but it updates the central template registry; the templates.go changes look syntactically incomplete in the diff (missing closing braces), which could break builds if not corrected.
Overview
Adds a new qa-agent TypeScript template that runs a Kernel-based visual QA workflow: it opens a Kernel browser session, uses Anthropic Computer Use to navigate/capture screenshots, then analyzes the final screenshot with a selectable vision provider (Claude/GPT-4o/Gemini) to produce structured issues plus JSON/HTML report generation helpers.
Wires the template into pkg/create/templates.go (metadata + default invoke command), adds template-specific docs/config (README.md, TESTING.md, env.example, package manifests/lockfiles), and updates .gitignore to ignore .pnpm-store.
Written by Cursor Bugbot for commit 41d08fe. This will update automatically on new commits. Configure here.