MCP Registry

unstructured-client

97557a49-e920-543e-9d1b-78a7d8e0a116 | Store v0.31.0
100/ 100
CRITICAL risk
AI review verdict
Do not install

The AI review rates the findings as likely false positive, but the risk score (100/100) still counts them.

Analysis record

Analysed
2 months ago 26 August 2026
Version
v0.31.0
Artifact
SHA256 4c677eb3…ac753e
Detail
Full list

Evidence ledger

Ranked by severity · findings with a source location link to the code viewer

324 findings
Showing 25 of 30 · highest severity first

YARA Rule Matches

18 rules
SeverityRuleHitsFilesMetadata
LOWRedirectToUnknownPath 2
bin/mcp-server.js.mapbin/mcp-server.js
-
LOWcredential env files 5
dist/esm/mcp-server/cli/start/impl.jsdist/commonjs/mcp-server/cli/start/impl.jsbin/mcp-server.js.map +2 more
-
LOWpostinstall persistence mechanism 7
dist/commonjs/sdk/models/shared/partitionparameters.d.tsdist/esm/sdk/models/shared/partitionparameters.d.tsdist/esm/sdk/models/shared/partitionparameters.js +4 more
-
LOWSurtrStrings 1
bin/mcp-server.js.map
-
LOWSurtr 1
bin/mcp-server.js.map
-
LOWDebuggerStatementsShouldNotBeUsed 1
bin/mcp-server.js.map
-
LOWpostinstall file download 62
dist/esm/lib/retries.d.tsdist/esm/mcp-server/tools/generalPartition.jsjsr.json +59 more
-
LOWNoUseWeakRandom 8
bin/mcp-server.jsdist/esm/hooks/custom/common.jsdist/esm/lib/retries.js +5 more
-
LOWNoUseEval 1
bin/mcp-server.js
-
LOWpostinstall obfuscation 60
dist/commonjs/funcs/generalPartition.d.tssrc/sdk/models/errors/index.tssrc/mcp-server/server.extensions.ts +57 more
-
LOWpostinstall file manipulation 31
dist/commonjs/lib/http.jsdist/commonjs/hooks/custom/FixArrayParamsHook.jsdist/commonjs/lib/primitives.js +28 more
-
LOWpostinstall environment access 14
src/mcp-server/server.extensions.tsdist/esm/mcp-server/server.extensions.d.tssrc/hooks/custom/HttpsCheckHook.ts +11 more
-
LOWNoUseSocketManually 2
bin/mcp-server.js.mapbin/mcp-server.js
-
LOWpostinstall registry modification 3
examples/package-lock.jsonbin/mcp-server.js.mapbin/mcp-server.js
-
LOWpostinstall network communication 65
dist/esm/lib/sdks.d.tstest/unit/FixArrayParamsHook.test.tsdist/esm/hooks/custom/LoggerHook.js +62 more
-
LOWpostinstall system command 27
dist/esm/lib/schemas.d.tsdist/commonjs/hooks/custom/LoggerHook.jsdist/commonjs/hooks/custom/LoggerHook.d.ts +24 more
-
LOWpostinstall crypto operations 2
bin/mcp-server.jsbin/mcp-server.js.map
-
LOWStaticallyServingHiddenFilesIsSecuritySensitive 2
bin/mcp-server.js.mapbin/mcp-server.js
-

Finding Categories

9
Secrets
4
Obfuscation
14
Network
163
IoC Indicators

YARA Rules Matched

18 rules(294 hits)
RedirectToUnknownPath credential env files postinstall persistence mechanism SurtrStrings Surtr DebuggerStatementsShouldNotBeUsed postinstall file download NoUseWeakRandom NoUseEval postinstall obfuscation postinstall file manipulation postinstall environment access NoUseSocketManually postinstall registry modification postinstall network communication postinstall system command +2 more

MCP Server Analysis

MCP servers expose tools and resources to AI assistants. Unlike browser extensions, they run as standalone processes with direct system access. Tool definitions are analyzed for prompt injection, data exfiltration, and tool poisoning patterns.

AI Security Report

AI Security Review

Evidence context: threat category none; evidence quality weak.

The unstructured-client package shows a critical risk score (100) primarily due to a massive volume of IoCs (1,037) and malware-signature findings (294). These findings are highly consistent with a bundled dependency false positive pattern. Specifically, the high IoC count suggests the package includes minified JavaScript or third-party libraries (dist/bundles) which trigger artifact scanners on generic URLs (CDNs) and code patterns.

The tool-poisoning findings (14) require careful calibration. In the context of an SDK or client library (as opposed to a standalone MCP server), these findings almost certainly represent tool definitions rather than tool poisoning. Legitimate libraries define tools with descriptions like "process file" or "extract data" for the AI to use. The scanner flags these descriptions, but they are intended for developer visibility and orchestration, not hidden manipulation. There is no evidence of invisible Unicode characters, "do not tell user" directives, or XML-based injection tags in this summary.

Without evidence of credential exfiltration (specific secret harvesting + unknown network C2) or confirmed shadow tool behavior, this package is classified as a likely false positive caused by scanner noise from bundled code.

Key Reasons

  • Findings consistent with bundled dependencies (high IoC/malware counts in client library)
  • Tool poisoning count (14) typical for SDKs defining tools vs. actual poisoning
  • No evidence of specific credential exfiltration (harvest + unknown C2)
  • Legitimate publisher name (Unstructured Technologies)

False Positive Considerations

  • Bundled dependencies in dist files
  • Tool definitions interpreted as tool poisoning
  • Generic IoC hits (CDNs, infra URLs)

Reviewed 2026-04-13; recommended action: suppress false positive; model confidence 70%.

Source Code Viewer

Source Code Not Available

Source code is not available for this version of the extension.

About This Extension

<h3 align="center"> <img src="https://raw.githubusercontent.com/Unstructured-IO/unstructured/main/img/unstructured_logo.png" height="200" > </h3>

Frequently Asked Questions