---
name: ping
description: >-
  Confirms an agent can reach the Skylark skills directory and read what it
  publishes. Use it as a first check when web skills are not loading, or to
  verify a new browser build discovers and activates a site skill at all.
runtime: browser
permissions:
  - page:read
scope:
  - https://skills.skylarkbrowser.com/*
triggers:
  - kind: url_pattern
    pattern: https://skills.skylarkbrowser.com/*
  - kind: intent
    description: check web skills are working
  - kind: intent
    description: ping the skills directory
version: 1.0.0
---

# Ping

An end to end check of the web skills pipeline. If you can follow these steps,
then discovery, fetching, digest verification, consent and activation have all
worked, so any problem lies somewhere else.

## Steps

1. Call `read_page`. You are on some page of
   `skills.skylarkbrowser.com`.

2. Find the page heading. Every page here has exactly one `h1`.

3. Report back, in one line:

   - that this skill activated,
   - the heading you found,
   - the current URL.

   For example: `ping: active. Heading "Agent skills for the sites you
   browse." at https://skills.skylarkbrowser.com/`.

4. Call `finish`. Do not click anything, do not navigate, and do not offer to
   do anything else unless the user asks.

## If you cannot read the page

Say so explicitly rather than guessing. A `read_page` that returns nothing on
this site means the page never rendered, which is a different problem from a
skill that failed to load. That distinction is the useful part of the report.

## What this proves

That the agent fetched an index from `/.well-known/agent-skills/index.json`,
downloaded a `SKILL.md`, checked its digest against what the index advertised,
surfaced it for consent, and injected it once the URL matched. Those are five
separate steps, and this skill activating means every one of them worked.