Back to skill

Security audit

元析 yotta-recon

Security checks for vulnerabilities and agentic risk

Overview

This network-scanning skill needs review because its authorization checks can allow probing beyond the approved scope.

Do not rely on Scope Guard as a complete network boundary until these authorization flaws are fixed. Use explicitly approved numeric IPs, a narrow scope file, --no-interactive, and avoid --assume-authorized. Pin a reviewed installation version and select only the intended agent. Both installer coverage gaps were manually investigated; validation used mocked connections, so live network behavior and registry package identity remain unverified.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/yotta_recon.py
Finding

DNS rebinding bypasses IP-based Scope Guard authorization

Content
View full analysis

Impact

TCP connections and service probes can reach out-of-scope public or internal addresses accessible from the scanner. The original successful scope check prevents the denied-target confirmation from being triggered. This establishes unauthorized probing, not arbitrary code execution or credential theft.

Vulnerability Patterns
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (22)

Scope Guard validates resolved addresses but scanning reconnects using the unchecked hostname.

High
Category
Excessive Agency
Confidence
99% confidence
Finding

Scope Guard checks hostname resolution at authorization time but passes the original hostname to later socket connections, which resolve it again. An attacker controlling DNS can initially return allowed addresses and subsequently return out-of-scope addresses, redirecting port scans and probes beyond the approved boundary; separate fingerprint connections create additional opportunities for this change.

Content

No source excerpt is available for this finding.

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

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 skills'.

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.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · README.md (reported line 42)May include surrounding context.

md
| Advantage | Description |
|---|---|
| **Zero dependency** | Python 3.8+ standard library; no daemon / database / external scanner; Windows + Linux + macOS |
| **Authorization discipline** | Only loopback is allowed by default; scope file or explicit declaration are the two authorization paths; in agent scenarios, no confirmation means denial |
| **Gentle & tunable** | Concurrency (--concurrency), timeout (--timeout) and rate limit (--rate) are adjustable to avoid scan storms |
| **Explainable fingerprints** | Product / version / risk level output item by item; risk hints match only, never exploit, and require human verification |
| **Flexible targets** | IP / CIDR / hostname / target file; custom ports or the built-in common-port table |

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
85% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.md (reported line 94)May include surrounding context.

text
# Optional China mirror: npm config set registry https://registry.npmmirror.com
npx -y @yottameta/yotta-recon --agent <agent-name>      # install to the agent's default user-level skills dir
npx -y @yottameta/yotta-recon --dir <your-skills-dir>   # point to the skills dir itself (e.g. ~/.codex/skills)

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

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The recommended installation command executes a registry-selected package version without pinning it, and '-y' suppresses npm's installation prompt. A compromised future release could execute with the user's privileges and install persistent agent instructions; this is an installation supply-chain weakness, not an MCP-specific issue.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The custom-directory installation example also executes an unpinned npm package with automatic confirmation. A malicious future package release could run arbitrary installer code and modify the selected agent's skill directory.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The upgrade guidance tells users to rerun an unpinned command to overwrite installed skill files with the latest release. This makes an upgrade an execution and persistent-instruction trust boundary that a compromised publisher account or future release could exploit.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The recommended command executes an unpinned npm package, with '-y' suppressing the installation prompt. A compromised future release could run with user privileges and install persistent agent instructions; the actual concern is installation supply-chain exposure rather than an MCP server.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The custom skill-directory example downloads and executes a package without selecting a reviewed version. If a future release is compromised, it could execute malicious installer code and place attacker-controlled instructions in that directory.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The upgrade instructions recommend executing the latest unpinned package and replacing installed skill content. A compromised future release could therefore gain code execution during installation and persist through the agent's skill-loading mechanism.

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, shell.

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-recon'.

Medium
Category
MCP Rug Pull
Confidence
88% confidence
Finding

The recommended installation command executes an unpinned npm package, while -y suppresses the installation confirmation. If users follow it, a compromised future release could execute with their permissions and install malicious skill content; this is documentation-level supply-chain exposure, not an MCP server configuration.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
88% confidence
Finding

This custom-directory installation example also executes an unpinned npm package without confirmation. A compromised release could execute arbitrary code before any installer destination checks, so specifying --dir does not mitigate the supply-chain risk.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
88% confidence
Finding

The global-install example executes an unpinned npm package and requests installation across known agent skill directories. If the selected release were compromised, this workflow could distribute malicious skill content across multiple agents, although the supplied code does not demonstrate such malicious behavior.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
88% confidence
Finding

The default installation example resolves and executes an unpinned package, potentially selecting an unreviewed future release. A compromised release could run with the invoking user's permissions and affect project-level agent skills; the comment itself does not execute anything.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
88% confidence
Finding

Although --list only prints directories in the supplied implementation, npx must execute the selected package to process that option. An unpinned compromised release could therefore execute malicious code even when the user only intends to inspect supported agents.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
88% confidence
Finding

The documented --dry-run option prevents installation writes in this implementation, but it does not prevent npx from executing the unpinned package. Users could mistakenly treat this as safe inspection, despite exposure to a compromised selected release.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
88% confidence
Finding

The version-query example still executes an unpinned npm package before displaying its version. If the selected release were compromised, the apparently informational command would provide the same execution opportunity as installation.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
88% confidence
Finding

The help example executes an unpinned package with automatic confirmation, even though the intended operation is informational. A compromised future release could execute arbitrary code before providing help; this is the same documentation-level supply-chain issue as the other examples.

Content

No source excerpt is available for this finding.

The documented authorization declaration plus confirmation requirement is not enforced.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

ScopeGuard.check_host immediately allows every target when 'assume_authorized' is set, without checking 'yes' or obtaining interactive confirmation. Consequently, '--assume-authorized' alone permits noninteractive scanning despite the documented declaration-plus-confirmation requirement, weakening the skill's authorization safeguard.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/yotta_recon.py (reported line 728)May include surrounding context.

python
out = []
    for cmd in (["netstat", "-ano"], ["ss", "-tln"]):
        try:
            proc = subprocess.run(cmd, capture_output=True, timeout=8,
                                  creationflags=getattr(subprocess, "CREATE_NO_WINDOW", 0))
            if proc.returncode != 0:
                continue