Vai al contenuto
TakoAPI
Torna alle skill

urlcheck

di cplusdev

Coding Agents & IDEs
1.1kDownload riportati da clawskills.sh
1Visualizzazioni
0 likeAccedi per mettere like a questa skill
—Non ancora valutato
La tua valutazione

Sintesi

Companion skill for @cybrlab/urlcheck-openclaw.

Descrizione

Cosa fa questa skill

Verifies URLs for security threats before an agent navigates, clicks, or follows a redirect. Accepts an optional intent parameter to check whether the destination matches what the user is actually trying to do, not just whether it's malicious.

Intent-alignment checking catches cases where a URL is technically safe but not the destination the user meant to reach, which threat-only scanners miss.

Quando usarla

  • Checking a login link before submitting credentials
  • Verifying a download URL before fetching a file
  • Screening redirects in automated browsing workflows
  • Confirming a payment or checkout page matches the user's intended site
  • Flagging phishing links in user-provided URLs

Esempio di workflow

INPUT

User asks: Before I open https://accounts-google.com.login-verify.net, check if it's safe to log in to my Google account

AGENT
  1. 1

    Call url_scanner_scan_with_intent with url='https://accounts-google.com.login-verify.net' and intent='log in to my Google account'

  2. 2

    Wait up to 90 seconds for the scan to complete

  3. 3

    Read the agent_access_directive field from the result

  4. 4

    Check intent_alignment to see if the destination matches a Google login page

  5. 5

    Report the directive, risk_score, and intent_alignment finding to the user

OUTPUT

DENY directive returned with high risk_score and intent_alignment: misaligned. Agent tells the user the URL does not appear to be a legitimate Google login page and recommends not proceeding.

Requisiti

  • openclaw plugins install @cybrlab/urlcheck-openclaw (free tier: 100 requests/day; contact contact@cybrlab.ai for higher limits)

Installazione

clawhub install cplusdev/urlcheck

Esegui il comando CLI oppure incolla il prompt in Claude Code, Cursor, Codex o nell'agente di coding che usi.

Link