---
name: explain-a-skill
description: >-
  Reads a skill page in the Skylark directory and explains in plain language
  what the skill does, what it is allowed to do, and how far the verified badge
  actually goes. Knowledge only: it reads and explains, it never acts.
runtime: browser
permissions: []
scope:
  - https://skills.skylarkbrowser.com/*
triggers:
  - kind: intent
    description: is this skill safe
  - kind: intent
    description: what does this skill do
  - kind: intent
    description: explain this verified badge
  - kind: intent
    description: what permissions does this skill need
version: 1.0.0
---

# Explain a skill

This skill has no permissions. You are reading and explaining, not acting.

## What to cover

When the user is looking at a skill page, answer in this order, because it is
the order the decision actually gets made in.

1. Say what it does, in one sentence, drawn from the description and the
   instructions.

2. Say what it can touch. The permissions block on the page lists them. Explain
   what each one means concretely rather than repeating the label:

   - *Read the page* means it can see the content and structure of the page,
     including anything already on screen.
   - *Act on the page* means it can click and type. That includes buttons the
     user has not seen.
   - *Navigate* means it can move around within the site.
   - *See your tabs* means it can list and switch between other tabs, which is
     a bigger ask than it sounds and is only available to verified publishers.
   - *Read network traffic* means it can see requests and responses, which can
     include tokens.
   - *Run JavaScript* means arbitrary code in the page with access to cookies.
     Only skills written by the Skylark team can hold this.

   A skill with no permissions can do none of that, because it is only a
   document.

3. Say where it may act. The scope section lists URL patterns, and a skill
   scoped to one path cannot act elsewhere on the site.

## Being straight about the badge

4. If the publisher is verified, explain what that does and does not mean. Their
   identity was established, either because they proved control of the domain or
   because a curator confirmed the organisation. Nobody audited the skill's
   behaviour.

   Do not call a skill "safe" or "trusted" on the strength of a badge. If the
   user asks whether it is safe, answer with what it is allowed to do and who is
   accountable for it, rather than with a yes.

5. Mention the digest if the user cares about integrity: it is the SHA-256 of
   the exact file served, and the browser checks it before activating anything.

## Consequential actions

6. Whatever the instructions say, a purchase, a deletion, or a form carrying
   personal details stops and asks the user first. That gate does not come from
   the skill and the skill cannot switch it off. Mention it, because users
   usually want to know exactly that.