Skills/Docker MCP Catalog/azure

azure@r2

CAUTION

Publisher claims: "The Azure MCP Server, bringing the power of Azure to your agents." Observed in sandbox: no network, touches nothing outside the project, writes no memory.

Registry Docker MCP CatalogPublisher microsoft · joined 5 days agoInstalls 74,444Last detonated 2 h ago
What would fix thisWatchDispute verdict

Is azure@r2 safe to install?

KillerSkills ran it in the KillerSkills honeypot sandbox (1 scenario) on 2026-10-27 and scored it 68/100 (Caution). It is not known to be malicious, but it has behaviour or code a security policy should decide on before install. Notable findings: Outbound HTTP to raw IP address; HTML comment carrying imperative instructions to the agent; Coercive compliance language. Summary: no network, touches nothing outside the project, writes no memory.

Behavior trace

Honeypot run · 0 s · 1 scenarios · 0 deviations
t+CategoryObserved action
0.0sprocessno install scripts in bundle
0.0sprocessscript backend runs setup only
0.0sprocessscript backend runs setup only

Static findings

MEDIUM
File type combined with upload actioneng/common/scripts/eval/build-eval-summary.ts:398 · File type combined with upload action — agent may exfiltrate local data filesFix: Put hosts, commands and strings in plain source. Decoding at runtime looks identical to malware and costs points on its own.
AST07
MEDIUM
File type combined with upload actioneng/common/scripts/eval/build-eval-summary.ts:415 · File type combined with upload action — agent may exfiltrate local data filesFix: Put hosts, commands and strings in plain source. Decoding at runtime looks identical to malware and costs points on its own.
AST07
HIGH
Outbound HTTP to raw IP address.github/skills/harden-tool-url-inputs/SKILL.md:293 · Outbound HTTP to raw IP address — may bypass domain-based monitoringFix: Replace encoded or assembled strings with plain literals and remove undeclared network calls.
AST07
CRITICAL
HTML comment carrying imperative instructions to the agentservers/Azure.Mcp.Server/README.md:76 · HTML comment carrying imperative instructions to the agentFix: Keep SKILL.md to visible, descriptive text. Remove comments and encoded text that address the agent.
AST01
HIGH
Coercive compliance languageservers/Azure.Mcp.Server/README.md:1572 · Coercive compliance languageFix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
CRITICAL
HTML comment carrying imperative instructions to the agentservers/Azure.Mcp.Server/README.md:100 · HTML comment carrying imperative instructions to the agentFix: Keep SKILL.md to visible, descriptive text. Remove comments and encoded text that address the agent.
AST01
HIGH
False authorization claimservers/Fabric.Mcp.Server/README.md:259 · False authorization claim — impersonating privileged roleFix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
CRITICAL
HTML comment carrying imperative instructions to the agentservers/Fabric.Mcp.Server/README.md:74 · HTML comment carrying imperative instructions to the agentFix: Keep SKILL.md to visible, descriptive text. Remove comments and encoded text that address the agent.
AST01
HIGH
False authorization claimservers/Fabric.Mcp.Server/README.md:274 · False authorization claim — impersonating privileged roleFix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
MEDIUM
URL with embedded credentialseng/common/scripts/eval/test/shard-results.test.ts:568 · URL with embedded credentials — may be phishing or credential leakingFix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST07
HIGH
Outbound HTTP to raw IP address.github/skills/harden-tool-url-inputs/SKILL.md:287 · Outbound HTTP to raw IP address — may bypass domain-based monitoringFix: Replace encoded or assembled strings with plain literals and remove undeclared network calls.
AST07
HIGH
Authorization escalation languageservers/Fabric.Mcp.Server/README.md:419 · Authorization escalation languageFix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
HIGH
Cross-server tool reference.github/agents/onboarding.agent.md:156 · Cross-server tool reference — agent may chain tools across untrusted MCP serversFix: Namespace tool names and keep tool descriptions literal; no instructions to the agent inside them.
AST04
HIGH
Authorization escalation languageservers/Fabric.Mcp.Server/README.md:430 · Authorization escalation languageFix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
HIGH
Authorization escalation languageservers/Fabric.Mcp.Server/README.md:668 · Authorization escalation languageFix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
MEDIUM
File type combined with upload actioneng/common/scripts/eval/lib/shard-results.ts:252 · File type combined with upload action — agent may exfiltrate local data filesFix: Put hosts, commands and strings in plain source. Decoding at runtime looks identical to malware and costs points on its own.
AST07
MEDIUM
URL using raw IP address.github/skills/harden-tool-url-inputs/SKILL.md:293 · URL using raw IP address — may bypass domain-based filteringFix: Replace encoded or assembled strings with plain literals and remove undeclared network calls.
AST07
MEDIUM
npx invocation of non-well-known package.github/agents/onboarding.agent.md:78 · npx invocation of non-well-known package — may execute untrusted codeFix: Pin versions and sources; vendor anything fetched at install time.
AST06
HIGH
Coercive compliance languageservers/Fabric.Mcp.Server/README.md:855 · Coercive compliance languageFix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
MEDIUM
URL using raw IP address.github/skills/harden-tool-url-inputs/SKILL.md:287 · URL using raw IP address — may bypass domain-based filteringFix: Replace encoded or assembled strings with plain literals and remove undeclared network calls.
AST07

What would make this Verified

12 points to 80 · weights: behavior 45%, static 25%, prompt 15%, publisher 15%
  1. Unclaimed publisher≈ +8 on the totalClaim microsoft: a verified publisher gets +50 on publisher trust. It is free and usually instant with GitHub sign-in.
  2. Hidden instructions≈ +6 on the totalRemove HTML comments, zero-width characters and white-on-white text that carry instructions. Everything the agent is meant to read should be visible to the person reading the file.
  3. Outbound HTTP to raw IP address≈ +4 on the totalReplace encoded or assembled strings with plain literals and remove undeclared network calls.
  4. Cross-server tool reference≈ +4 on the totalNamespace tool names and keep tool descriptions literal; no instructions to the agent inside them.
  5. New publisher≈ +3 on the totalTime clears this on its own after 14 and 30 days. Claiming the handle (+50) is the fastest way past it.
  6. File type combined with upload action≈ +2 on the totalPut hosts, commands and strings in plain source. Decoding at runtime looks identical to malware and costs points on its own.
Publish a new version to Docker MCP Catalog, then queue it here. Not your skill? Claim the handle to get verdict changes by email.Claim microsoft