Skip to content
Squatch MCP

Findings

Detected by Squatch rules from the server's metadata and protocol behavior. Latest completed scan: .

Severity: High4 findings

Severity: HighStatus: OpenSQ-TOOL-EXFIL-DIRECTIVE

Asks the AI model to pass conversation data or secrets

Model-facing text instructs the AI model to place conversation content, prompts, tool history or secrets into tool arguments or external destinations.

First seen · Last seen

Suggested fix: Do not ask the model to include conversation, prompt or credential content in arguments.

Severity: HighStatus: OpenSQ-TOOL-EXFIL-DIRECTIVE

Asks the AI model to pass conversation data or secrets

Model-facing text instructs the AI model to place conversation content, prompts, tool history or secrets into tool arguments or external destinations.

First seen · Last seen

Suggested fix: Do not ask the model to include conversation, prompt or credential content in arguments.

Severity: HighStatus: OpenSQ-TOOL-EXFIL-DIRECTIVE

Asks the AI model to pass conversation data or secrets

Model-facing text instructs the AI model to place conversation content, prompts, tool history or secrets into tool arguments or external destinations.

First seen · Last seen

Suggested fix: Do not ask the model to include conversation, prompt or credential content in arguments.

Severity: HighStatus: OpenSQ-TOOL-EXFIL-DIRECTIVE

Asks the AI model to pass conversation data or secrets

Model-facing text instructs the AI model to place conversation content, prompts, tool history or secrets into tool arguments or external destinations.

First seen · Last seen

Suggested fix: Do not ask the model to include conversation, prompt or credential content in arguments.

Severity: Medium2 findings

Severity: MediumStatus: OpenSQ-SERVER-AUTH-NONE-WRITE

State-changing tools available without authentication

The remote server lists state-changing tools without requiring authentication.

First seen · Last seen

Suggested fix: Require MCP authorization for servers that expose state-changing tools.

Severity: MediumStatus: OpenSQ-TOOL-INJECTION-INSTRUCTIONS

Instructions addressed to the AI model

Model-facing text contains instructions addressed to the AI model rather than a description of the tool (for example to override prior instructions, to act before using tools, or to react to trigger phrases).

First seen · Last seen

Suggested fix: Describe what the tool does and its parameters. Do not include instructions to the model about other tools, the user or the conversation.

Severity: Low2 findings

Severity: LowStatus: OpenSQ-TOOL-ANNOTATION-UNDERCLAIM

Annotations understate what the tool does

The tool declares itself read-only, but its name or description indicates it changes, deletes or runs something. Clients that auto-approve read-only tools would run it without confirmation.

First seen · Last seen

Suggested fix: Set readOnlyHint to false and destructiveHint to reflect the tool's behavior.

Severity: LowStatus: OpenSQ-REPO-MISSING

No source repository linked

No source repository is linked in the registry entry.

First seen · Last seen

Suggested fix: Add a repository to server.json.

4 resolved or dismissed findings
Severity: HighStatus: ResolvedSQ-TOOL-EXFIL-DIRECTIVE

Asks the AI model to pass conversation data or secrets

Model-facing text instructs the AI model to place conversation content, prompts, tool history or secrets into tool arguments or external destinations.

First seen · Last seen

Suggested fix: Do not ask the model to include conversation, prompt or credential content in arguments.

Status history
  1. ResolvedSquatch
Severity: HighStatus: ResolvedSQ-TOOL-EXFIL-DIRECTIVE

Asks the AI model to pass conversation data or secrets

Model-facing text instructs the AI model to place conversation content, prompts, tool history or secrets into tool arguments or external destinations.

First seen · Last seen

Suggested fix: Do not ask the model to include conversation, prompt or credential content in arguments.

Status history
  1. ResolvedSquatch
Severity: MediumStatus: ResolvedSQ-TOOL-INJECTION-INSTRUCTIONS

Instructions addressed to the AI model

Model-facing text contains instructions addressed to the AI model rather than a description of the tool (for example to override prior instructions, to act before using tools, or to react to trigger phrases).

First seen · Last seen

Suggested fix: Describe what the tool does and its parameters. Do not include instructions to the model about other tools, the user or the conversation.

Status history
  1. ResolvedSquatch
Severity: MediumStatus: ResolvedSQ-TOOL-INJECTION-INSTRUCTIONS

Instructions addressed to the AI model

Model-facing text contains instructions addressed to the AI model rather than a description of the tool (for example to override prior instructions, to act before using tools, or to react to trigger phrases).

First seen · Last seen

Suggested fix: Describe what the tool does and its parameters. Do not include instructions to the model about other tools, the user or the conversation.

Status history
  1. ResolvedSquatch

What this scan does not cover

  • Squatch never calls a server's tools, so it cannot see what they actually do at run time.
  • It scans without credentials: tools behind authentication may not be visible.
  • It does not review source code, dependencies or the publisher's infrastructure.
  • A server can change behavior between scans; monitoring compares each scan with the last.
  • Classifications come from tool names, descriptions and schemas, which the publisher controls.

How Squatch scans servers