Back to skill

Security audit

元案 yotta-lesson

Security checks for vulnerabilities and agentic risk

Overview

The skill coherently supports local lesson planning, with installation precautions and a timing-check limitation users should understand.

Use a reviewed, fixed release and select the intended agent or directory. Preview installation locally with --dry-run and back up customized existing skill files. Manually verify lesson timings: a passing check does not validate the sum of stage durations. The bundled installers were reviewed, but an npm release retrieved by an unpinned command may differ.

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
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (16)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The visible commands introduce a network-capable package execution step that is inconsistent with the declared no-network operation. The --agent and --dir arguments suggest setup or integration, but do not establish that any model is called. This finding is limited to the supplied invocation: it does not demonstrate that the installed tool itself accesses the network, nor does it verify the advertised lesson-planning functionality.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The supplied chunk implements package deployment rather than lesson generation or deterministic lesson checking. Its substantive filesystem installation and cleanup capabilities are not described in the declared purpose, and the chunk contains no implementation of the advertised lesson workflow or findings gate. Installation can support the broader skill, but this executable's primary behavior is materially different from the supplied description. The code uses Node.js built-ins and shows no network access or model calls, so those local-only claims are consistent. This assessment is limited to the supplied chunk and does not establish whether the advertised functionality exists elsewhere.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

Although this is an installer rather than an MCP server, the recommended command can retrieve and execute an unpinned npm release with the user's privileges. A compromised future release could modify agent skills or access local files; the lesson engine's offline operation does not protect the installation step.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The custom-directory installation example executes an npm package without selecting a fixed version. Although not an MCP configuration, it exposes users to changed or compromised package code before any installation-directory restriction could help.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The recommended installation command can execute a changing npm release with the user's privileges. This is an installer, not an MCP server, but the underlying supply-chain concern is valid and could affect persistent agent skills despite the offline lesson-processing design.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

This command leaves the executed npm package version unspecified. A compromised release would execute before the installer handles the custom destination, so choosing a destination does not constrain malicious package code.

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-lesson'.

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

This comment is a usage example, not code that automatically runs npx, but following it can execute an unpinned package release. The MCP label is inaccurate; the real concern is supply-chain execution with access to user files and persistent agent skill directories.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

Following this custom-destination example can retrieve and execute an unspecified npm release. It is not an MCP server reference, but a malicious release would run with the user's privileges regardless of the requested destination.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The documented global installation command executes an unpinned npm release. A compromised release could access local data or distribute altered skills across multiple agents, making this installation context more consequential than ordinary lesson processing.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

The bare npx example leaves the package release mutable and executes it if followed. Although this is documentation for an installer rather than an MCP server, a compromised release could modify project-level agent skills or other writable files.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
92% confidence
Finding

The list-only example still asks npx to resolve and execute an unpinned package before the package interprets --list. The supplied installer only prints directories in this mode, but a compromised resolved release need not honor that limitation.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

The dry-run flag does not prevent npx from retrieving and executing an unpinned package. The shown installer respects dry-run, but a compromised release could perform writes or access data before or instead of honoring that flag.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
92% confidence
Finding

Even the version-query example can execute a newly resolved, unpinned npm release. The current code's harmless version branch does not constrain a compromised future package, although the comment itself performs no execution.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
92% confidence
Finding

Requesting help through this example still executes an unpinned npm package if the command is followed. This is not an MCP configuration, but compromised package code could run before help is displayed and is not confined by the educational purpose.

Content

No source excerpt is available for this finding.

The advertised timing-gap checks do not validate individual stage timings or their actual sum.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The checker accepts any minute annotation and only compares an optional stated total with the expected duration; it never sums actual stage durations or requires timing for each stage. A malformed or deliberately padded lesson can therefore pass CI with missing or inconsistent timings, undermining the advertised integrity check, though the impact is limited to lesson-quality validation.

Content

No source excerpt is available for this finding.