ワンクリックで
browse
Use gstack's persistent browser in OpenCode for screenshots, QA flows, console checks, and browser-driven verification.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
メニュー
Use gstack's persistent browser in OpenCode for screenshots, QA flows, console checks, and browser-driven verification.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Check local gstack setup health, detect stale build/setup state, and write a bounded upgrade-readiness report.
Interrogate an idea like a bounded YC-style office hours session and write a durable local memo.
Review recent local repository history and write a bounded engineering retrospective.
Prepare a local authenticated browser session for OpenCode QA and design workflows using gstack browse cookie import flows.
Audit a local rendered page with browser evidence, write a bounded design report, and optionally apply local UI fixes when explicitly requested.
Update a small, factual set of project docs from local changes and write a durable summary without committing or using GitHub APIs.
| name | browse |
| description | Use gstack's persistent browser in OpenCode for screenshots, QA flows, console checks, and browser-driven verification. |
| compatibility | opencode |
| metadata | {"host":"opencode","migration_phase":"1"} |
This skill gives OpenCode agents a fast persistent browser using gstack's local browse binary.
Use it when you need to:
The browser is stateful. Cookies, tabs, and local storage persist across commands.
Before any browser action, confirm the binary path:
if [ -x "./browse/dist/browse" ]; then
B="./browse/dist/browse"
elif command -v browse >/dev/null 2>&1; then
B="$(command -v browse)"
else
echo "NEEDS_SETUP"
fi
If the result is NEEDS_SETUP, run the repo setup first:
./setup
If you need to do it manually instead, build it from this repo:
bun install && bun run build && bunx playwright install chromium
Then re-run the check and continue.
Prefer the repo-local binary at ./browse/dist/browse when it exists.
snapshot -i when you need to discover page elementssnapshot -D after important interactions to confirm what changedconsole after interactions, not just after the first page loadsnapshot after navigation because refs become stale$B goto https://example.com
$B text
$B console
$B network
$B goto https://example.com/login
$B snapshot -i
$B fill @e3 "user@example.com"
$B fill @e4 "password"
$B click @e5
$B snapshot -D
$B snapshot -i -a -o /tmp/annotated.png
$B screenshot /tmp/page.png
$B console
$B responsive /tmp/layout
$B viewport 375x812
$B screenshot /tmp/mobile.png
$B diff https://staging.example.com https://prod.example.com
Use snapshot as the primary inspection tool.
snapshot -i for interactive refssnapshot -a -o <file> for annotated screenshotssnapshot -D to see what changed after an actionsnapshot -C when a UI uses clickable divs or custom components that do not show up cleanly in the accessibility treeWhen you write screenshots or annotated images, use the file read path afterward so the result is visible to the user. A screenshot that only exists on disk is not enough evidence.