Back to skill

Security audit

元呈 yotta-present

Security checks for vulnerabilities and agentic risk

Overview

Review is needed because chart input can overwrite local files, and automatic formatting can block or rewrite replies.

Before enabling this globally, address the chart-input file-write behavior: reject embedded out/filename fields or limit filesystem write access to an approved output folder. Approve saved settings separately and make companion rewriting opt-in. Use an exact reviewed package version. Installer coverage gaps were investigated directly; companion behavior and actual host hook enforcement remain 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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Taint TrackingDirect Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (21)

MCP server referenced without pinned version: 'npx one-liner'.

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-present'.

Medium
Category
MCP Rug Pull
Confidence
97% confidence
Finding

This recommended installation command executes an npm package without selecting a reviewed version, and -y suppresses the installation confirmation. A compromised future release could execute with the user's privileges and modify agent skill files; the renderer's offline operation does not protect this installation step.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
97% confidence
Finding

The custom-directory installation example executes whichever package release the registry currently resolves, rather than an approved artifact. A malicious release would run with the caller's privileges, not merely the permissions implied by the requested destination directory.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
97% confidence
Finding

Although --list appears informational, npx still downloads and executes the unpinned package to implement it. A compromised release could therefore execute arbitrary code even when the user only intends to inspect supported installation directories.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
97% confidence
Finding

The recommended command executes an unpinned npm installer with confirmation suppressed by -y. Compromise of a subsequently resolved release could grant code execution under the user's account and access to agent skill files, regardless of the installed renderer's offline design.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
97% confidence
Finding

This example resolves and executes a mutable npm release to install into a user-selected directory. The directory argument does not sandbox the installer, so a compromised package could access other files available to the calling user.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
97% confidence
Finding

The --list option still requires executing the npm package, and no reviewed version is selected. A malicious future release could run arbitrary code during what appears to be a read-only discovery operation.

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, file_write.

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.

The workflow automatically delegates prose rewriting despite the manifest's presentation-only boundary.

Medium
Category
Excessive Agency
Confidence
95% confidence
Finding

The workflow instructs the agent to send rendered prose to an installed rewriting skill based on its own judgment, without requiring authorization for that transformation. This crosses the stated presentation-only boundary and can alter qualifications, warnings, or conclusions after rendering, without a specified post-rewrite fidelity check; merely having the companion installed is not consent to use it on every eligible response.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
96% confidence
Finding

The optional companion's installation command executes an unpinned npm package with -y. The instruction not to install it automatically reduces immediate exposure, but a user following the recommendation could still execute a compromised future release with their account's privileges.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
86% confidence
Finding

The recommended installation command uses an unpinned npm package with automatic confirmation. If a user follows it after a compromised release is published, attacker-controlled code could execute with the user's permissions and install persistent agent instructions; however, this is a documented installer command, not an MCP server launch or an automatically executed statement in this file.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
86% confidence
Finding

The custom-directory installation example recommends executing an unpinned npm package with automatic confirmation. A compromised release could execute code with the user's permissions regardless of the selected installation directory; this risk arises only if the documented command is followed, and the MCP-specific classification is inaccurate.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
86% confidence
Finding

The global installation example executes an unpinned npm package, making the recommended workflow dependent on the release resolved at execution time. A compromised release could execute arbitrary user-level code, while the intended global installation also distributes skill content across multiple agent directories; this is documentation rather than an executable MCP configuration.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
86% confidence
Finding

The default installation example recommends running an unpinned npm package with automatic confirmation. Following it could execute a compromised release and place untrusted skill instructions in detected project agent directories; the comment itself does not execute anything, and no MCP launch is shown.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
86% confidence
Finding

Although the example only requests a directory listing, npx must still resolve and execute the unpinned package. A compromised release need not honor the --list option, so following this documentation can expose the user to arbitrary code execution; this is not evidence of an actual MCP server invocation.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
86% confidence
Finding

The --dry-run option prevents installation writes in the supplied implementation, but it does not prevent npx from executing the unpinned package. A compromised release could ignore dry-run behavior and execute arbitrary code, so the documented command retains supply-chain risk despite appearing non-mutating.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
86% confidence
Finding

Requesting --version through npx still executes the unpinned package rather than safely querying registry metadata. If a compromised release is resolved, it could execute arbitrary user-level code before or instead of printing a version; this is a conditional documentation-level risk, not an MCP configuration issue.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
86% confidence
Finding

The help example executes an unpinned npm package with automatic confirmation, even though users may consider help inspection harmless. A compromised release could execute arbitrary code without honoring --help; the supplied file only documents this command and does not launch an MCP server.

Content

No source excerpt is available for this finding.

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

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.

Tainted flow: 'path' from sys.stdin.read (line 1909, user input) → open (file write)

Medium
Category
Data Flow
Confidence
65% confidence
Finding

Data from a source is assigned to a variable that is later passed to a sink, creating a variable-mediated taint flow.

Content

Scanner excerpt · scripts/yotta_present.py (reported line 1846)May include surrounding context.

python
def _write_utf8(path, text):
    with open(path, "w", encoding="utf-8", newline="\n") as f:
        f.write(text)

The automatic rendering hook has an overly broad, open-ended activation condition.

Medium
Category
Trigger Abuse
Confidence
92% confidence
Finding

The hook applies to an open-ended category of structured final messages and specifies blocking when the required renderer is not used. On hosts that enforce it, this can turn an optional presentation tool into a dependency for unrelated responses, causing availability or output-policy conflicts; this manifest does not encode an opt-in requirement or explicit exemptions.

Content

No source excerpt is available for this finding.