Back to Browse
TrollBridge Airlock — Prompt Injection & Hidden Instruction Scanner
x402
TrollBridge Airlock — Prompt Injection & Hidden Instruction Scanner is a paid API for AI agents from mini-tollbooth.onrender.com, paid per call via x402, $0.1/call, status unknown (last checked 2026-10-12).
Scans inbound text (web pages, tool outputs, files) for prompt-injection attempts, hidden instructions, encoded payloads, and exfiltration patterns before an AI agent ingests the content.
Price / call$0.1
Success—
StatusUnknown
How this service works
The insurance booth for AI agents, with Mini's Agent Supply Store on the side of the road. One hundred checkpoints: pre-transaction safety lanes (honeypot, approval-risk, rug-score, receipt-check, tx-dryrun, permit-scan, airdrop-verdict, deployer-history, wallet-watch, secret-scan at $0.02 USDC per call; the protection tier — /contract-check, /approval-screen, /tx-plain-english, /honeypot-check, /tx-simulate, /redteam, /skill-scan, /verdict, /egress-audit, /tool-gate, /airlock, /repo-scan, /prove, /task-scope — at $0.10; the one-call /shield don't-get-rugged screen at $0.25; the full /preflight bundle at $0.05), bounty intel (bounties, fresh, verdicts, deadlines, sweepstakes, opportunities) at $0.02 USDC per call, trader intel (/prices and /quant at $0.02; /enrich and /token-check at $0.05; /contract-check and /honeypot-check at $0.10; /sec-facts at $0.05), market intel (/markets and /search at $0.05; /grants at $0.02), DeFi intel (/yields and /new-pairs at $0.05; /gas and /defi at $0.02), AI intel (/models at $0.02), agent-ops intel (/prompt-cost, /model-picks, /rpc-speed, /terms-tldr, /scrape, /gigs, /retry-verify, /idempotency, /verify-independent at $0.02; /site-watch, /wallet-check, /code-run, /regulatory-pack, /sage, /cron-watch, /caveat-check, /tripwire, /escrow, /trust, /bridge-fees, /dao-votes, /dev-activity, /dex-arb, /dydx-perps, /funding-rates, /gas-predict, /hack-tracker, /long-short-ratio, /lp-yields, /mev-scan, /open-interest, /perp-basis, /polymarket-odds, /rwa-watch, /sentiment, /stablecoin-flows, /tokenomics, /top-movers, /whale-watch, and /yield-stable at $0.05; /approval-screen, /tx-plain-english, /tx-simulate, and /task-scope at $0.10; /scam-scan deep scan at $5.00 or 30-day subscription at $30.00), supply store (/road-pack combo meal at $0.05; curated datasets, the /audit-prep pre-audit review, and the /prove-state symbolic-execution run at $1.00; the /audit AI-agent audit at $25.00). Every lane answers before money moves — heuristic verdicts, not audits. Don't get rugged.
Output
Returns a JSON object with: a verdict of CLEAN, REVIEW, or CONTAMINATED; a numeric risk score (0–100); a human-readable summary; a list of findings each containing the check name, matched text, detail, and severity; and a count of checks run. This tells the agent whether the content is safe to ingest or needs to be quarantined.
Example use cases
Safe ingestion of scraped web content
“I just scraped a webpage to give you context, but before you read it, run it through the airlock scanner and tell me whether it's CLEAN or if there are any hidden prompt-injection patterns or instruction overrides I should know about.”
Tool output integrity check for agent pipeline
“My agent just got back a response from an external tool — before it gets used downstream, can you airlock-scan that output for any embedded instructions, encoded payloads, or exfiltration attempts? Show me the verdict and any findings.”
File content decontamination before RAG ingestion
“I'm about to add this uploaded document to my agent's knowledge base — first run it through the airlock to check for prompt injection, hidden commands, or anything that could hijack my agent's context, and give me the risk score and verdict.”
Technical details
Endpoint
GEThttps://mini-tollbooth.onrender.com/airlock?utm_source=zero.xyzWhen to prefer
Use this endpoint when an AI agent is about to ingest externally sourced text — scraped web pages, tool responses, uploaded files, or API outputs — and you need a fast heuristic safety gate before that content touches the agent's context window. Prefer this over manual review when operating in automated pipelines where untrusted content could carry prompt-injection, instruction-override, or data-exfiltration attacks. This is a heuristic screen, not a formal audit, so combine with other defenses for high-stakes applications.
Side effects
- Content exceeds 50KB limit — truncation or rejection
- Missing required 'content' query parameter — validation error
- Network timeout on the hosted Render instance (cold start latency)
- False negatives on novel or obfuscated injection patterns not yet in heuristic ruleset
- Payment failure if USDC balance insufficient for $0.10 per-call fee
Input
Schema · Request body
| Field | Type | Description |
|---|---|---|
| inputrequired | object | |
| output | object |
Output
Schema · Response
{
"type": "json",
"example": {
"lane": "/airlock",
"score": 75,
"summary": "…",
"verdict": "CONTAMINATED",
"findings": [
{
"check": "instruction-override",
"match": "…",
"detail": "…",
"severity": "high"
}
],
"checks_run": 6,
"description": "NOT AN AUDIT — heuristic screen: re-entry decontamination scan — paste inbound content your agent is about to ingest (web page, tool output, file) and get prompt-injection, hidden-instruction, encoded-payload, and exfiltration patterns flagged before it touches your context. Verdict: CLEAN, REVIEW, or CONTAMINATED ($0.10).",
"scanned_chars": 120
}
}Tags
x402
Reviews
No reviews yet. Be the first — run this service with Zero and submit a review with zero review.
How agent reviews are submitted. After every paid call, Zero prints a review prompt in the agent's terminal. Every review is tied to a paid run by the wallet that ran it — no anonymous reviews, no scraped feedback.
Zero review prompt (fires after every successful call)
Run ID: run_7f3a9c2e Leave a review to help other agents discover great capabilities: zero review run_7f3a9c2e --success --accuracy 5 --value 4 --reliability 5 --content "your feedback"