← Back to Discover

LochBot

Browser tool that checks a pasted chatbot system prompt against 31 prompt-injection patterns and returns a 0-100 security score with fix suggestions. The analysis runs in the page.

AI security & guardrails
Visit website ↗

What it does

LochBot is a free web page where you paste the system prompt of a chatbot and get back a review of how well its wording defends against known prompt-injection tactics. It checks 31 attack patterns grouped into seven categories, including direct injection, delimiter attacks, data extraction, role-play jailbreaks and encoding attacks. Results are scored from 0 to 100 with a letter grade and a severity for each weakness, and every failed check comes with a suggested line to add to the prompt. The report can be exported as JSON. The checker does not send your prompt to a model. The site says it uses local JavaScript pattern matching in the browser with no API calls, and describes itself as a first-pass structural review rather than a behavioural red-team test against a running model.

Capabilities

  • 31 attack patterns across 7 categories
  • 0-100 security score with a letter grade and severity levels
  • Specific fix suggestions for each failed check
  • Runs client-side in the browser with no API calls
  • JSON export of the vulnerability report

When to use it

  • You are reviewing a chatbot system prompt before it goes live.
  • You want to compare how defensive two prompt drafts are.
  • You want a quick structural check before doing real red-team testing.