Skip to content
verify mcp Beta VerifyMCP is currently in beta. If you notice any issues, get in touch and we’ll put it right.

Prompt Sanitizer — prompt-injection scan for untrusted content

REMOTE · ANGEL.OROBOROS-DESIGN.COM

Scan web pages, mails, PDFs and tool output for prompt injection before your agent reads them.

+2 this week 65 Trust /100

Last scanned (20 hours ago)

Scores higher than 46% of remote servers

Reachable. Last reached 21 Nov 2026.

Category Security

Trust breakdown (7 categories)

How this component scores in each security and reliability category. Every signal is checked automatically against the live server, and we only credit what we can confirm. How we score → Why this is hard to score →

Endpoint Security63
Transport & Reachability100
Schema Quality & AI Usability60
  • AI-judged instruction clarity (good).Pass
  • Context-footprint check failed: tool/resource definitions use about 940 tokens (~235/item across 4 items; 4 tools + 0 resources), over budget; trim descriptions and params. See how to fix → Fail
  • Usage-examples check failed: none of the tools include examples. See how to fix → Fail
Stability & Change Management23
  • Stability observed for 7 of 30 days with no destabilising changes; credit accrues until the full window elapses.Partial
Tool Coverage92
  • 100% of tools have a non-trivial description (not blank, and not just the tool's name).Pass
  • 77% of tool parameters carry a description.Partial
Tool Safety75
  • No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.Pass
  • 0 of 1 tool(s) whose name or description implies an irreversible operation declare an MCP destructiveHint annotation; "input_guard" implies "remove" and declares readOnlyHint instead, contradicting what its own name says it does. See how to fix → Fail
  • An AI judge read all 5 captured unit(s) of tool text and found none that tries to manipulate the model reading it.Pass
Capabilities100
  • Implements a supported MCP spec version (2025-11-25); the latest is 2026-07-28.Pass
What we tested

Checks we ran (12)

  • TLS certificate
  • HTTPS
  • HSTS header
  • DNSSEC
  • Transport
  • Schema quality
  • Stability
  • Tool coverage
  • Injection markers
  • Destructive annotations
  • Judged manipulation
  • MCP protocol

Checks we could not complete (1)

These are gaps on our side, not findings about this hosted endpoint. We only credit what we can confirm, so a category made up entirely of them scores 0.

  • Authorization in Endpoint SecurityAuthorisation not fully verified: no authorisation is required to call this server, and 2 tool(s) never declared a destructiveHint. The MCP spec treats an absent hint as destructive by default, so we cannot call this surface safe.

Does not apply to this hosted endpoint (2)

  • Supply Chain SecurityScored on package channels only. This is a hosted endpoint, so there is nothing to install or inspect.
  • Provenance & TransparencyScored on package channels only. This is a hosted endpoint, so there is nothing to install or inspect.

The score does not measure everything. What the score does not measure →

Changelog

Every change we have recorded for this component, newest first. Security-relevant changes are always shown. ▲ marks a change for the better, ▼ a change for the worse; unmarked changes are neutral.

  • 20 Nov 26 +1

    No change was recorded against any check on this day. Stability & Change Management went from 17 to 20. That category is still filling its 30-day observation window: 5 days of observed history at the previous scan, 6 at this one. The score rises as the window fills, whether or not the server changes.

  • 19 Nov 26 −1
    • The server rewrote its instructions, which are the text every model session reads security
    • Tool “buy_pack_with_card” rewrote its description, which is the text the model reads security
    • Tool “input_guard” rewrote its description, which is the text the model reads security
    • Tool “stealth_read” rewrote its description, which is the text the model reads security
    • Destructive annotations: pass → 0 functional
    • “buy_pack_with_card” reworded the description of “email” cosmetic
    • “input_guard” reworded the description of “bundle_token” cosmetic
    • “input_guard” reworded the description of “payer_address” cosmetic
    • “stealth_read” reworded the description of “bundle_token” cosmetic
    • “stealth_read” reworded the description of “payer_address” cosmetic
    • Tool “buy_pack_with_card” changed its title: TradeSignal: pack by card → Angel: 5,000 calls by card cosmetic
    • Tool “input_guard” changed its title: Prompt-injection scan for untrusted web, email and tool content → Remove hidden content from untrusted text and flag text aimed at the model cosmetic
    • Tool “stealth_read” changed its title: Read any URL as clean, injection-scrubbed Markdown for an LLM → Read any public page as clean Markdown, with hidden text removed cosmetic
  • 17 Nov 26 +1
    • The server rewrote its instructions, which are the text every model session reads security
  • 16 Nov 26 0
    • The server rewrote its instructions, which are the text every model session reads security
    • Tool “buy_pack_with_card” rewrote its description, which is the text the model reads security
    • Schema quality: 664 → 967 ▼ functional
    • Tool coverage: 67% → 77% ▲ functional
    • New tool “stealth_read” functional
  • 15 Nov 26 +1
    • The server rewrote its instructions, which are the text every model session reads security
    • Tool “input_guard” rewrote its description, which is the text the model reads security
    • Stability: unverified → 0.03 ▲ functional
  • 14 Nov 26 63

    First indexed and scored.

Diagnostics

Diagnostic detail from the automated scan of this channel: what the scanner observed at each step, so you can see exactly where a check passed or failed. It is informational only and never changes the trust score.

Captured 21 Nov 2026 · Probed https://angel.oroboros-design.com/mcp/prompt-sanitizer

TLS valid

Negotiated TLS 1.3 with TLS_AES_128_GCM_SHA256 .

Subject Issuer Valid from Valid until Key Signature Serial
CN=*.oroboros-design.com CN=YR2,O=Let's Encrypt,C=US 28 Oct 2026 26 Jan 2027 RSA 2048 SHA256-RSA 5c407b368cf93bd9f2f03db9f58a77f6c9f
SANs: *.oroboros-design.com, oroboros-design.com
CN=YR2,O=Let's Encrypt,C=US (CA) CN=Root YR,O=ISRG,C=US 3 Sept 2025 2 Sept 2028 RSA 2048 SHA256-RSA 4ebd24947e24d394802d84a52fd5b319
CN=Root YR,O=ISRG,C=US (CA) CN=ISRG Root X1,O=Internet Security Research Group,C=US 13 May 2026 2 Sept 2032 RSA 4096 SHA256-RSA f24b6d17f9d9ad7cb1c9fea78782699f

Background: What to check on a remote MCP endpoint →

DNSSEC insecure

Validation of angel.oroboros-design.com. — Not signed

Zone DS Keys Algorithms Outcome
. Trust anchor 20326, 38696 8, 8 Verified
com. DS published 19718 13 Verified
oroboros-design.com. No DS Unsigned (proven) parent-signed NSEC/NSEC3 proves an unsigned delegation
Authentication No authorisation required

The endpoint answered without asking for a token. Anyone who knows the URL can reach it.

Result No authorisation required
HTTP status 200
Header Value
strict-transport-security max-age=63072000

Background: How OAuth 2.1 works in the 2026 MCP spec →

Transports 2 probes
Transport URL Outcome Status Location
streamable-http https://angel.oroboros-design.com/mcp/prompt-sanitizer Verified 200
http (plaintext) http://angel.oroboros-design.com/mcp/prompt-sanitizer HTTPS enforced 308 https://angel.oroboros-design.com/mcp/prompt-sanitizer
MCP tools · 4 exposed · ~816 tokens

The tools this component advertises to a client, with an estimated token cost for each. Expand a tool to see its parameters and schema. The per-tool counts are indicative and are not scored directly; the schema's total context footprint is one signal in Schema Quality & AI Usability. A tool's description is untrusted text the model reads on every call, which is what makes this list a security surface and not just an inventory: how tool poisoning works →

Tool Tokens
buy_pack_with_card ~118

For businesses without USDC: 5,000 calls €9.00, one-time, 90 days, businesses only, paid in EUR by card or Link. Returns checkout_url for a person to open; the checkout asks for the company or freelance name. After payment the success page (Accept: application/json) returns the key, and it is e-mailed. After: any tool with bundle_token.

NameTypeReqDescription
callsinteger––
emailstring–Where the key is sent (optional; Stripe asks otherwise).

No output schema declared.

No examples provided.

input_guard ~318

Removes hidden content and flags text aimed at the model, for untrusted input before an agent acts on it: web pages, emails, RAG chunks, tool and MCP responses. Strips invisible Unicode (tag smuggling decoded, zero-width, bidi) and HTML hidden from people; flags role takeover, forged system or tool turns, requests for keys, the prompt or payments, exfiltration links, homoglyphs and base64 payloads, in 17 languages. Rule-based. Verdict clean, caution or hostile, score, findings, cleaned text. Free with a key (500 calls) or 5 calls a day without one; then 0.02 USDC per call via x402, or 1 call from a key. After: preflight before your agent pays anything the content pointed to (the result's next field has the exact call).

NameTypeReqDescription
bundle_tokenstring–Your key instead of paying per call: free (500 calls, POST /key) or bought by card. The connection header X-Bundle-Token works too.
payer_addressstring–Wallet you pay from (optional; the x402 price is the same for every address).
payment_txstring–Hash of a USDC transfer on Base to payTo (≥ the price, ≤ 72 h old) instead of an x402 signature.
textstring–the content to check (send text or url)
urlstring–a public https page to fetch once and check (256 KB)

No output schema declared.

No examples provided.

input_guard_preview ~105

Free prompt-injection quick scan for untrusted input: up to 4,000 characters of text, or one https page, checked for hidden instructions aimed at agents (role override, invisible Unicode, hidden HTML) in 17 languages incl. Chinese, Japanese, Korean, Hindi, Arabic. Returns verdict, score and the top 3 findings. After: input_guard for every finding and the cleaned text.

NameTypeReqDescription
textstring––
urlstring––

No output schema declared.

No examples provided.

stealth_read ~275

Returns a public https page as clean Markdown for an agent (headings, lists, tables kept; navigation, cookie banners, scripts and everything hidden from people removed), strips invisible Unicode and flags text aimed at the model — with verdict, score, findings, title and language. JSON and text pass through. One GET as a declared bot, no login or bot-protection bypass; a blocked page is never charged. Free with a key (500 calls) or 5 calls a day without one; then 0.02 USDC per call via x402, or 1 call from a key. After: preflight before your agent buys or pays anything the page offers (the result's next field has the exact call).

NameTypeReqDescription
bundle_tokenstring–Your key instead of paying per call: free (500 calls, POST /key) or bought by card. The connection header X-Bundle-Token works too.
payer_addressstring–Wallet you pay from (optional; the x402 price is the same for every address).
payment_txstring–Hash of a USDC transfer on Base to payTo (≥ the price, ≤ 72 h old) instead of an x402 signature.
urlstringyesthe public https page to read (1 MB, 10 s)

No output schema declared.

No examples provided.

Install

How do I install the Prompt Sanitizer — prompt-injection scan for untrusted… MCP server?

Prompt Sanitizer — prompt-injection scan for untrusted… is a hosted endpoint at https://angel.oroboros-design.com/mcp/prompt-sanitizer, so there is nothing to install locally. Ready-made configuration for Claude, Cursor, VS Code, Codex and 5 more is on this page, copied from each client's own documentation.

remote · angel.oroboros-design.com

# add to Claude Code
claude mcp add --transport http com-oroboros-design-angel-prompt-sanitizer 'https://angel.oroboros-design.com/mcp/prompt-sanitizer'
Common questions

What is the Prompt Sanitizer — prompt-injection scan for untrusted… MCP server?

Prompt Sanitizer — prompt-injection scan for untrusted… is an MCP server listed in the public MCP registry as com.oroboros-design.angel/prompt-sanitizer. Scan web pages, mails, PDFs and tool output for prompt injection before your agent reads them. This page covers its hosted endpoint (https://angel.oroboros-design.com/mcp/prompt-sanitizer).

Is the Prompt Sanitizer — prompt-injection scan for untrusted… MCP server safe to use?

Prompt Sanitizer — prompt-injection scan for untrusted… scores 65 out of 100 on VerifyMCP. That is a record of what we were able to check automatically, not an endorsement. The category breakdown on this page shows every signal behind the number, including the ones we could not confirm.

What tools does the Prompt Sanitizer — prompt-injection scan for untrusted… MCP server expose?

Prompt Sanitizer — prompt-injection scan for untrusted… exposes 4 tools: stealth_read, input_guard, buy_pack_with_card, input_guard_preview. Their descriptions and schemas cost roughly 816 tokens of context every time the server is loaded.

Does the Prompt Sanitizer — prompt-injection scan for untrusted… MCP server require authentication?

No. We connected to Prompt Sanitizer — prompt-injection scan for untrusted… without credentials and it answered, so anything it exposes is reachable by anyone who knows the address.

Is the Prompt Sanitizer — prompt-injection scan for untrusted… MCP server still maintained?

Prompt Sanitizer — prompt-injection scan for untrusted… is still listed as active in the MCP registry. We last reached this channel on 21 November 2026. Those dates come from our own scans of the registry and the channel itself, not from anything the publisher announced.

What did VerifyMCP test on the Prompt Sanitizer — prompt-injection scan for untrusted… MCP server?

VerifyMCP ran 12 checks on Prompt Sanitizer — prompt-injection scan for untrusted…, covering TLS certificate, HTTPS, HSTS header, DNSSEC and 8 more. One item could not be completed, a gap on our side rather than a finding about Prompt Sanitizer — prompt-injection scan for untrusted…: Authorization. Supply Chain Security and Provenance & Transparency do not apply to a hosted endpoint, so they were not tested. The score does not measure everything, and the scoring documentation lists what it leaves out.