shiplightai/agent-skills
6 skills2.9k installs12 starsGitHub
Install
npx skills add https://github.com/shiplightai/agent-skillsSkills in this repo
1Verifyverify is a Shiplight agent skill that validates UI changes in a real browser through Shiplight MCP tools. The skill defines a 10-step workflow: build a verification plan, start the dev server, open a session with new_session, read the DOM via inspect_page before screenshots, interact with act, check get_browser_console_logs, close_session, and optionally generate_html_report or upload_html_report for shareable evidence. A daily cache check refreshes skills at most once per 24 hours using .shiplight-agent-skills-last-update. Developers reach for verify after layout, styling, component, or navigation changes and skip it for backend-only, docs-only, or test-only edits. Authenticated apps can save_storage_state to ~/.shiplight/ for reused login sessions across runs.571installs2Cloudcloud is a Shiplight agent skill for bidirectional sync between a local repo and Shiplight cloud. Developers push YAML test cases, reusable templates, and TypeScript test functions upward, then pull cloud definitions back down so the same cases run on a schedule, integrate with CI pipelines, and stay shared across a team. Setup requires a Shiplight cloud subscription and a SHIPLIGHT_API_TOKEN; MCP tools handle most sync workflows while the REST API manages test runs, environments, folders, suites, and accounts. Reach for cloud when browser-automation tests outgrow local-only runs and need centralized scheduling, environment targeting, or team-wide test libraries without rewriting cases in a separate UI.537installs3Reviewreview is the Shiplight agent-skills orchestrator and single entry point for application reviews. It triages context, then loads domain playbooks from references/ for security (OWASP, auth, supply chain), privacy (PII, GDPR/CCPA), compliance (HIPAA, SOC 2, PCI-DSS), design (responsive, a11y, i18n), resilience (errors, API contracts), performance (Core Web Vitals, bundles), SEO (meta, structured data), and GEO (llms.txt, AI citation readiness). Three modes cover default triage (/review), full suite (/review --all), and targeted domain runs (/review security --quick). Use review for pre-launch readiness, post-incident planning, or any request to review an app when the needed domain is unclear.530installs4Triagetriage is a shiplightai agent skill for repairing failing Shiplight YAML tests without masking real application bugs. The six-step workflow reproduces the narrowest failing target, reads failure output plus matching specs under specs/tests/ and knowledge/ notes, inspects the live app when logs are insufficient, applies the smallest correct fix to YAML templates helpers or auth setup, validates edits with validate_yaml_test, reruns the changed target, and updates specs or knowledge/ when durable lessons emerge. Ground-truth precedence ranks explicit user instructions and feature specs above current app behavior—agents must report mismatches instead of rewriting test intent to match broken UI. Skip triage for net-new test authoring (/create-yaml-tests) or passing-test quality passes. Reach for triage when CI Shiplight runs fail after a deployment, several tests share one broken locator, or a UI change broke existing YAML coverage.524installs5Create Testscreate-tests is a Claude Code skill for testing & qa. It helps solo builders move faster with AI-assisted development.437installs6Cloud V2cloud_v2 is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted coding.319installs