imyesobo/my-agent-skills
Portable Agent Plugin: a Playwright record/author/replay toolkit for real sites with logins and 2FA, an Azure DevOps intake skill, a session-improve review skill, plus a plan-with-Opus implement-with-Sonnet delegating agent for Copilot CLI.
Retrieve Azure DevOps work item information, including details, relations, history, and linked items. Use when analyzing a work item before implementation or debugging Azure DevOps state.
Read a Figma design (layout, styles, variables/tokens, components, images, comments) and its Code Connect mappings using the Figma REST API and Code Connect CLI, without the Figma MCP. Use when given a figma.com URL or when implementing UI from a Figma design.
Drive a headed browser with playwright-cli to record a user journey on any website step by step, including shadow-DOM / web-component sites. Use when: open a headed browser, playwright-cli open/snapshot/click/fill/run-code/eval, keep a running list of '### Ran Playwright code', wait for the user to log in or pass 2FA/captcha, state-save .auth/state.json, an element is not clickable or has no accessible name, intercepts pointer events, strict mode violation, find the clickable ancestor.
Turn a recorded running list of playwright-cli steps into a Playwright spec tests/<name>.spec.ts that loads storageState, contains no login or credentials, reuses or creates a per-site helpers file (manual-login wait, overlay dismissal, re-auth wait, downloads, popups, native print stub) and asserts every step. Use when: write the spec, generate the test file from the running list, add a section to an existing spec, which helper to use, what assertion to add after a step, create playwright.config.ts.
Run a Playwright spec headed, coordinate manual login / 2FA with the user, read the failure (log, error-context.md, trace) and apply the matching fix pattern, then re-run until it passes. Use when: replay the spec, run the test, npx playwright test, the script is stuck, test failed, strict mode violation, intercepts pointer events, waiting for event popup, Login required, Security validation required, element not found after goBack.
Orchestrates recording, authoring and replaying a Playwright end-to-end spec for any website: drive a headed browser with playwright-cli step by step together with the user, write tests/<name>.spec.ts from the recorded steps, then replay it until it passes. Use when asked to record a test, explore a site in a browser with me, build an e2e test for a page, create the spec from what we just did, or replay an existing spec. Trigger phrases: record a test, drive the browser, playwright-cli open --headed, write the spec from the running list, replay the spec, shadow DOM locators, the user will log in.
Review the session that just ended and propose the changes to this plugin (skills, agents, workflows) that would have spent fewer AI credits on the next similar session. Use when the user asks to run session-improve, to review the session, or to make the next run cheaper.