agentleFS
saffron-healthsaffron-health/opencode-gui/.agents/skills/dev-command/SKILL.md

dev-command

Skill59 starsChanged 9 months agoMITView on GitHub

Interact with the running VSCode extension via Playwright. Use when automating, testing, or debugging the OpenCode webview UI.

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    All selectors below target elements inside frame.

    line 29
  2. Very specific:

    Do NOT pass undefined as a second argument to frame.evaluate(() => ...) — it causes "Too many arguments" errors.

    line 107

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Dev Command

Launch and interact with the VSCode extension using pnpm dev.

Commands

pnpm dev                    # Launch VSCode in background tmux session
pnpm dev exec "<code>"      # Execute JS in the webview (frame/page/browser available)
pnpm dev snapshot           # Screenshot the VSCode window
pnpm dev stop               # Stop the session

Exec environment

The exec command provides three objects:

  • frame — the webview active-frame (SolidJS UI). Use this for most interactions.
  • page — the top-level VSCode Electron page. Use for screenshots or VSCode-level actions.
  • browser — the Playwright Browser instance.

Common selectors

All selectors below target elements inside frame.

Input

# Type a message
pnpm dev exec "await frame.locator('[contenteditable]').pressSequentially('hello world')"

# Submit (Cmd+Enter)
pnpm dev exec "await frame.locator('[contenteditable]').press('Meta+Enter')"

# Clear and retype
pnpm dev exec "await frame.locator('[contenteditable]').fill(''); await frame.locator('[contenteditable]').pressSequentially('new text')"

Buttons

# Stop generation
pnpm dev exec "await frame.locator('.shortcut-button--stop').click()"

# Submit button
pnpm dev exec "await frame.locator('.shortcut-button--secondary').click()"

# New session
pnpm dev exec "await frame.locator('.new-session-button').click()"

# Session switcher
pnpm dev exec "await frame.locator('.session-switcher-button').click()"

Permissions

# Allow once
pnpm dev exec "await frame.locator('.permission-button--primary').click()"

# Allow always
pnpm dev exec "await frame.locator('button[aria-label=\"Allow always\"]').click()"

# Reject
pnpm dev exec "await frame.locator('button[aria-label=\"Reject\"]').click()"

# Allow all visible permissions
pnpm dev exec "const btns = await frame.locator('.permission-button--primary').all(); for (const b of btns) await b.click()"

Messages

# Get all message texts
pnpm dev exec "const msgs = await frame.locator('.message-text').allTextContents(); return msgs"

# Get last assistant message
pnpm dev exec "const msgs = await frame.locator('.message--assistant .message-text').allTextContents(); return msgs[msgs.length - 1]"

# Wait for assistant response
pnpm dev exec "await frame.locator('.message--assistant').last().waitFor()"

# Check if thinking
pnpm dev exec "return await frame.locator('.loading-indicator').isVisible()"

Reading state

# Get input text
pnpm dev exec "return await frame.locator('[contenteditable]').textContent()"

# Count messages
pnpm dev exec "return await frame.locator('.message').count()"

# Get session error
pnpm dev exec "return await frame.locator('.session-error').textContent()"

Important notes

  • Do NOT pass undefined as a second argument to frame.evaluate(() => ...) — it causes "Too many arguments" errors. Just call with one arg.
  • Screenshots must be taken from page, not frame — CDP screenshot only works on top-level targets.
  • The webview frame takes ~10-15s to appear after VSCode starts. The exec command handles this automatically.
  • The input uses Tiptap (ProseMirror), so use pressSequentially instead of fill for realistic typing. fill works for clearing.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.