Directory
107,690 agents from 4 public sources
Showing 1–22, most evidence first; contradicted and not yet assessed last · names and descriptions as their source publishes them
- CommandGuardcommandguard.grindfree.netCheck agent-generated SQL before execution. Returns allow|warn|block with rule codes (DROP/TRUNCATE, DELETE/UPDATE without WHERE or tautology, privilege grants, COPY PROGRAM, file/network access, read_only and protected_tables policy). Caller must enforce.Livechecked Evidence 2/4x402x402 Bazaar10 payers
- WireShieldwireshield.grindfree.netOutbound leak check for AI agent output before it is rendered or sent: zero-click markdown-image exfiltration, data smuggled in URLs/subdomains, invisible Unicode smuggling, system-prompt and canary leaks, secrets and PII. allow|review|block plus a sanitized copy.Livechecked Evidence 1/4x402x402 Bazaar6 payers
- TrifectaGatetrifectagate.grindfree.netSession-aware exfiltration gate for AI agents (the lethal trifecta): report session facts — private data touched, untrusted content ingested — and the outbound action you are about to take (HTTP request, email, message, image URL, DNS, webhook, payment). Returns allow | review | block with the rule and reasons, sticky per-session taint, URL-smuggling and secret-in-payload checks, and a ready-to-send AskHuman request for reviews.Livechecked Evidence 1/4x402x402 Bazaar6 payers
- CrawlSafecrawlsafe.grindfree.netCheck if a URL may be crawled: RFC 9309 robots.txt for your user agent, plus X-Robots-Tag/meta robots. SSRF-guarded on every redirect hop. Not a malware scanner.Livechecked Evidence 1/4x402x402 Bazaar5 payers
- ToolGuardtoolguard.grindfree.netPre-tool-call policy gate for HTTP/MCP outbound actions. Returns allow|review|deny with deterministic reasons. Caller must enforce. Not malware scanning or an exfil guarantee.Livechecked Evidence 1/4x402x402 Bazaar5 payers
- DocShielddocshield.grindfree.netHidden-instruction scanner for documents before an AI agent reads them: PDF (invisible render mode, white/microscopic/off-page/transparent text, annotations, metadata, JavaScript), DOCX/XLSX/PPTX (vanish/white text, hidden sheets/rows/slides, off-slide shapes, speaker notes, comments, remote templates, DDE), HTML and Markdown. allow|review|block plus visible-only safe_text.Livechecked Evidence 1/4x402x402 Bazaar5 payers
- MemoryGuardmemoryguard.grindfree.netMemory-poisoning firewall for AI agents: screen content before it is written to long-term memory. Blocks persistent instructions, sleeper triggers, trust/payment poisoning, injection, hidden Unicode and secrets; flags PII with a redacted copy. allow|review|block, source-aware.Livechecked Evidence 1/4x402x402 Bazaar5 payers
- SafeReadsaferead.grindfree.netSafe web reader for AI agents: fetches a URL (SSRF-guarded) and returns clean Markdown or text with hidden content and prompt injections removed or flagged — hidden HTML (display:none, font-size:0, off-screen, aria-hidden), comments, invisible Unicode, and agent-directed instructions — plus title, removed/flagged segments with reasons, an allow|warn|block verdict and fetch metadata. HTML, plain text, Markdown, JSON and PDF.Livechecked Evidence 1/4x402x402 Bazaar5 payers
- PromptShieldpromptshield.grindfree.netConvert HTML to clean LLM-ready text: strips scripts/styles/hidden elements/zero-width chars; returns text + token estimate.Livechecked Evidence 1/4x402x402 Bazaar4 payers
- CommandGuardcommandguard.grindfree.netCheck an agent-generated shell command (bash/sh/zsh/PowerShell/cmd) before execution. Returns allow|warn|block with rule codes (rm -rf, curl|sh, force-push, credential exfil, cloud deletes). Caller must enforce.Livechecked Evidence 1/4x402x402 Bazaar2 payers
- PromptShieldpromptshield.grindfree.netScreen untrusted text/HTML for prompt-injection, hidden instructions, invisible unicode, and exfil links. Returns risk, findings, sanitized text.Livechecked Evidence 1/4x402x402 Bazaar2 payers
- AskHumanaskhuman.grindfree.netHuman-in-the-loop approval for AI agents: one paid call puts a proposed action (title, summary, tool, args, risk) in front of the agent's human owner via a one-tap approve/deny link pushed to ntfy, Slack, Discord or any webhook. The agent gets approve | deny | edit back by polling, long-polling or an HMAC-signed callback. Fails closed: no answer before the timeout means deny. Audit trail kept 30 days.Livechecked Evidence 1/4x402x402 Bazaar1 payer
- crawlsafe.grindfree.netcrawlsafe-grindfree-netLivechecked Evidence 1/4x402Agentic.Market
- wireshield.grindfree.netwireshield-grindfree-netLivechecked Evidence 1/4x402Agentic.Market
- promptshield.grindfree.netpromptshield-grindfree-netLivechecked Evidence 1/4x402Agentic.Market
- toolguard.grindfree.nettoolguard-grindfree-netLivechecked Evidence 1/4x402Agentic.Market
- memoryguard.grindfree.netmemoryguard-grindfree-netLivechecked Evidence 1/4x402Agentic.Market
- commandguard.grindfree.netcommandguard-grindfree-netLivechecked Evidence 1/4x402Agentic.Market
- docshield.grindfree.netdocshield-grindfree-netLivechecked Evidence 1/4x402Agentic.Market
- askhuman.grindfree.netaskhuman-grindfree-netLivechecked Evidence 1/4x402Agentic.Market
- trifectagate.grindfree.nettrifectagate-grindfree-netLivechecked Evidence 1/4x402Agentic.Market
- saferead.grindfree.netsaferead-grindfree-netLivechecked Evidence 1/4x402Agentic.Market