From mayring-coder
Autonomously walk through the whole logged-in web app (app.linn.games), screenshot every page, and surface bugs. Use when the user wants a full UI sweep, an end-to-end check after changes, "durchspielen / durchklicken", or to hunt for broken pages, console/network errors, or visual glitches across the app. Read-only and safe — never submits, deletes, pays, or logs out.
How this skill is triggered — by the user, by Claude, or both
Slash command
/mayring-coder:app-walkthroughThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Drives a headless Chromium (Playwright) through the app, capturing per page:
Drives a headless Chromium (Playwright) through the app, capturing per page: screenshot · JS-console errors · network responses ≥400 · HTTP status · Laravel/Livewire error pages. Writes a JSON report + PNGs that you then read and analyse — the script flags technical errors; you catch visual/semantic bugs (duplicate cards, empty states, broken layout) by reading the screenshots.
Safety invariant: read-only. Only GET-navigation + at most accepting the cookie banner. A deny-list blocks anything matching logout/delete/pay/submit/ save/export. No state change, no money, no logout. Do NOT relax this.
WV="$HOME/.cache/app-walkthrough/venv"
SCRIPT="$(ls "$HOME"/.claude/plugins/cache/*/mayring-coder/*/bin/app_walkthrough.py 2>/dev/null | sort | tail -1)"
Bootstrap once (skip if $WV/bin/playwright exists):
python3 -m venv "$WV" && "$WV/bin/pip" install -q playwright \
&& "$WV/bin/playwright" install chromium chromium-headless-shell
Ensure a session (skip if ~/.cache/app-walkthrough/state.json exists and a
later walkthrough doesn't print NOT_LOGGED_IN). The login is headful — it
opens a browser window; tell the user to sign in via GitHub, the script
detects success and persists the session for all future headless runs:
DISPLAY="${DISPLAY:-:1}" "$WV/bin/python" "$SCRIPT" --login
Run this in the background and wait for LOGIN_OK.
Walk (headless, no interaction):
"$WV/bin/python" "$SCRIPT" --out /tmp/app-walkthrough
Run in the background; it prints WALK_DONE: N pages, M with findings.
Analyse /tmp/app-walkthrough/report.json:
error, console_errors, network_errors,
dom_flags, or status>=400 are non-empty.--base URL — target a different host (default https://app.linn.games).--routes /a,/b — override the seed route list (link-discovery still adds more).--out DIR — output directory (default /tmp/app-walkthrough).If a walkthrough prints NOT_LOGGED_IN, the persisted session expired — re-run
step 2.
npx claudepluginhub nileneb/mayringcoder --plugin mayring-coderProvides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.