guard inspect

Screen untrusted content (web pages, emails, tool output) for prompt-injection/manipulation before your agent trusts it. Returns a risk score + reasons; a scree

liveunknown~5s typical

mcpfetcher

$0.0020

per call

plus $0.001 platform fee per run

Open MCP server; Cracked bills routing only

Provider Upstream docs

$Set up https://cracked-ai-pearl.vercel.app/SKILL.md, then use Cracked to run mcp-cloud-fetcher-fetcher /guard_inspect for me.

Input

Field	Type	Description
text*	string	Untrusted content to screen.
deep	boolean	Use the ML classifier too (default true).

Call it

curl

curl https://cracked-ai-pearl.vercel.app/v1/run \
  -H "Authorization: Bearer ck_live_..." -H "content-type: application/json" \
  -d '{"provider":"mcp-cloud-fetcher-fetcher","endpoint":"/guard_inspect","input":{"text":"..."}}'

cli

npx cracked-ai run -p mcp-cloud-fetcher-fetcher -e /guard_inspect -i '{"text":"..."}'

mcp

run_tool({ provider: "mcp-cloud-fetcher-fetcher", endpoint: "/guard_inspect", input: {"text":"..."} })

Try it

Runs against your signed-in workspace balance. Sign in if you have not.