Back to skill

Security audit

元镜 yotta-mirror

Security checks for vulnerabilities and agentic risk

Overview

This local analysis tool needs review because its sharing guidance overstates how well it protects student identities.

Treat student reports and error logs as sensitive even with --anonymize. Before sharing, use independently assigned random student codes or remove student-level details, and inspect the result. Manually review low-sample conclusions. Prefer reviewed local installation files, preview one destination with --dry-run, and back up customized installations. The bundled files were inspected, but any release subsequently downloaded by the unpinned npm command remains unverified.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (19)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The visible commands run a materially different tool from the one described. The -y option permits npx to obtain the package without an installation confirmation when it is unavailable locally. Although the package's internal behavior is not supplied and network access may be unnecessary if it is cached, the commands do not implement the declared offline diagnostics functionality. There is no evidence here establishing model calls, data exfiltration, or the package's precise effects.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The commands use npx -y, which permits npm to obtain and execute an external package without an installation confirmation. This introduces an external dependency and potential network access despite explicit claims of zero dependencies and no networking. The visible commands concern agent/directory configuration rather than score processing or diagnostic reporting. The package implementation is not supplied, so its downstream behavior—including whether it calls models—cannot be determined from this chunk.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
90% confidence
Finding

The supplied chunk performs deployment and installation maintenance rather than educational analysis. Ordinary copying could be a supporting implementation detail, but the cross-agent installation scope and destructive cleanup of existing installations are material filesystem behaviors not represented in the declared purpose. No network access or model calls are evident, consistent with those restrictions. This finding concerns the supplied installer; it does not establish whether other files implement the advertised analytics.

Content

No source excerpt is available for this finding.

Anonymized reports are promoted for sharing without warning that hashed student identifiers remain linkable and potentially re-identifiable.

Medium
Category
Output Handling
Confidence
99% confidence
Finding

The documented unsalted SHA-256 prefix is pseudonymization, not anonymization: the same identifier produces the same code across datasets, and likely names or student numbers can be tested offline. Promoting these reports as shareable and not reversible across datasets can cause disclosure of identifiable student performance information.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y @yottameta/yotta-mirror'.

Medium
Category
MCP Rug Pull
Confidence
91% confidence
Finding

Although this is an installer rather than an MCP server, the command executes an unpinned npm package and automatically accepts installation. A compromised future release could execute with the user's privileges, potentially accessing locally stored student records or modifying persistent agent skills.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y @yottameta/yotta-mirror'.

Medium
Category
MCP Rug Pull
Confidence
91% confidence
Finding

The explicit-directory installation command executes an unpinned npm package with automatic confirmation. Although not an MCP invocation, it exposes the installation environment to malicious future package releases, including unauthorized file access or persistent skill modifications.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y @yottameta/yotta-mirror'.

Medium
Category
MCP Rug Pull
Confidence
91% confidence
Finding

The recommended command automatically executes an unpinned npm installer, rather than an MCP server. If a subsequent release is compromised, installation could expose local student data or introduce persistent malicious agent instructions.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y @yottameta/yotta-mirror'.

Medium
Category
MCP Rug Pull
Confidence
91% confidence
Finding

Specifying a destination directory does not constrain the privileges of the unpinned package executed by npx. Automatic installation of a compromised release could therefore affect files beyond that destination and modify persistent agent behavior.

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: env, file_read.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y @yottameta/yotta-mirror'.

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y @yottameta/yotta-mirror'.

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y @yottameta/yotta-mirror'.

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y @yottameta/yotta-mirror'.

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y @yottameta/yotta-mirror'.

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y @yottameta/yotta-mirror'.

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y @yottameta/yotta-mirror'.

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y @yottameta/yotta-mirror'.

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Anonymization mode still exposes original student identifiers in validation errors.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

Even when anonymization is enabled, duplicate-identifier, invalid-number, and out-of-range errors interpolate the original student identifier. These messages reach stderr and can expose student identities through CI logs or shared troubleshooting output despite the user's privacy setting.

Content

No source excerpt is available for this finding.

Low-sample knowledge assessments remain definitive despite the documented mandatory manual-review downgrade.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

Knowledge levels are assigned without checking the minimum respondent threshold, and those definitive labels are retained in JSON and Markdown. Later logic only suppresses low-sample weak diagnoses, leaving other low-sample knowledge conclusions without equivalent knowledge-level review status and allowing consumers to bypass the promised reliability safeguard.

Content

No source excerpt is available for this finding.