docs(#61, #95): add timeout configuration and improve multilingual skill triggers#472
docs(#61, #95): add timeout configuration and improve multilingual skill triggers#472giulio-leone wants to merge 2 commits intovercel-labs:mainfrom
Conversation
|
@giulio-leone is attempting to deploy a commit to the Vercel Labs Team on Vercel. A member of the Team first needs to authorize it. |
Replace incorrect timeout configuration section with accurate wait command guidance. The timeout field exists per-command in the JSON protocol but has no CLI flag or environment variable. Addresses Vercel Agent Review feedback on PR vercel-labs#472. Co-authored-by: Copilot <[email protected]>
Replace incorrect timeout configuration section with accurate wait command guidance. The timeout field exists per-command in the JSON protocol but has no CLI flag or environment variable. Addresses Vercel Agent Review feedback on PR vercel-labs#472.
4fc999a to
02f95c1
Compare
Replace incorrect timeout configuration section with accurate wait command guidance. The timeout field exists per-command in the JSON protocol but has no CLI flag or environment variable. Addresses Vercel Agent Review feedback on PR vercel-labs#472.
d3e555d to
35d8648
Compare
There was a problem hiding this comment.
Pull request overview
Updates the agent-browser skill metadata to improve auto-discovery/triggering (including multilingual phrases) and adjusts the allowed tool invocation pattern.
Changes:
- Expand the skill frontmatter
descriptionwith additional English trigger phrases plus multilingual equivalents. - Restrict
allowed-toolstoBash(agent-browser:*)(removing thenpx agent-browser:*fallback).
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
skills/agent-browser/SKILL.md
Outdated
| name: agent-browser | ||
| description: Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction. | ||
| allowed-tools: Bash(npx agent-browser:*), Bash(agent-browser:*) | ||
| description: Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", "browse the web", "check a webpage", "navigate to a URL", "web scraping", "download page content", or any task requiring programmatic web interaction. Also triggers on multilingual equivalents: "ouvrir un site web", "abrir una página", "aprire un sito", "Webseite öffnen", "ウェブサイトを開く", "打开网站", "웹사이트 열기", "открыть сайт". |
There was a problem hiding this comment.
PR description mentions documenting an AGENT_BROWSER_TIMEOUT env var, but the codebase/docs use AGENT_BROWSER_DEFAULT_TIMEOUT (and there are no references to AGENT_BROWSER_TIMEOUT). Consider updating the PR description (or the documentation text, if intended) so the env var name is consistent with the actual implementation.
There was a problem hiding this comment.
Fixed — the SKILL.md uses the correct AGENT_BROWSER_DEFAULT_TIMEOUT env var name, consistent with the codebase. Branch rebased onto main and pushed.
skills/agent-browser/SKILL.md
Outdated
| name: agent-browser | ||
| description: Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction. | ||
| allowed-tools: Bash(npx agent-browser:*), Bash(agent-browser:*) | ||
| description: Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", "browse the web", "check a webpage", "navigate to a URL", "web scraping", "download page content", or any task requiring programmatic web interaction. Also triggers on multilingual equivalents: "ouvrir un site web", "abrir una página", "aprire un sito", "Webseite öffnen", "ウェブサイトを開く", "打开网站", "웹사이트 열기", "открыть сайт". |
There was a problem hiding this comment.
The frontmatter description is very long and likely exceeds the intended metadata budget. The repo’s skill authoring guidance notes metadata is always in context and should be ~100 words; consider shortening this to a compact keyword/phrase list (and move longer trigger guidance into the SKILL body if needed) to avoid context bloat.
There was a problem hiding this comment.
Fixed — trimmed the frontmatter description to ~89 words by removing verbose trigger phrases and replacing individual multilingual examples with a summary reference. Branch rebased onto main and pushed.
skills/agent-browser/SKILL.md
Outdated
| description: Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction. | ||
| allowed-tools: Bash(npx agent-browser:*), Bash(agent-browser:*) | ||
| description: Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", "browse the web", "check a webpage", "navigate to a URL", "web scraping", "download page content", or any task requiring programmatic web interaction. Also triggers on multilingual equivalents: "ouvrir un site web", "abrir una página", "aprire un sito", "Webseite öffnen", "ウェブサイトを開く", "打开网站", "웹사이트 열기", "открыть сайт". | ||
| allowed-tools: Bash(agent-browser:*) |
There was a problem hiding this comment.
allowed-tools previously permitted both agent-browser:* and npx agent-browser:* (matching other skills like dogfood/slack and the README’s “Node.js fallback”). Dropping the npx form may prevent this skill from working in environments where the binary isn’t globally installed; consider keeping Bash(npx agent-browser:*) as a fallback even if the text recommends preferring the native binary for performance.
| allowed-tools: Bash(agent-browser:*) | |
| allowed-tools: Bash(agent-browser:*), Bash(npx agent-browser:*) |
There was a problem hiding this comment.
Fixed — re-added Bash(npx agent-browser:*) to allowed-tools as a fallback for environments without a global install. Branch rebased onto main and pushed.
…mprove multilingual skill triggers
33d9ca8 to
db49f6e
Compare
Summary
Improves SKILL.md with timeout configuration documentation and enhanced trigger sensitivity with multilingual support.
Changes
AGENT_BROWSER_TIMEOUTenv var and wait strategies for slow websitesCloses #61, closes #95