Back to skill

Security audit

dialogue_graph

Security checks for vulnerabilities and agentic risk

Overview

This skill is a small dialogue-graph helper library with optional visualization and no evidence of hidden, destructive, persistent, or credential-related behavior.

Install optional Graphviz dependencies only in a trusted, isolated environment, and consider pinning graphviz before use. The skill otherwise appears purpose-aligned and low risk.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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)

T08 · Insecure Dependencies

Note
Location
SKILL.md:62
Finding

Unpinned Third-Party Graphviz Dependency

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, line 62
Vulnerability Type: Unpinned third-party dependency
Risk Level: Low

Code Snippet:

python
# Requires: pip install graphviz

Technical Analysis

The documentation instructs users to install the graphviz Python package without specifying a reviewed version or verifying an integrity hash. Consequently, the package content resolved at installation time may differ from the version originally reviewed.

This does not establish that the named package is malicious. The risk arises from mutable dependency resolution: a compromised package release, package-index account, configured index, or network/package-source environment could cause unreviewed code to be installed. No lock file, version constraint, hash, or trusted-index requirement is provided to reduce this supply-chain exposure.

Attack Path

  1. An attacker compromises a relevant package release or the package source configured in the victim environment.
  2. A user or Agent follows the documented pip install graphviz instruction.
  3. pip resolves and downloads the currently available package rather than a specifically reviewed artifact.
  4. Malicious package installation logic or runtime code executes with the privileges of the account performing the installation or importing the package.
  5. The malicious dependency can access resources available to that account and affect subsequent visualization operations.

Impact Assessment

Exploitation is conditional on compromise or manipulation of the dependency supply chain; no direct compromise exists solely from the documented command. If exploited, code could execute with the installing user's privileges. The accessible scope could include project files, environment variables, credentials readable by that account, and network resources available to the process. The instruction does not itself request elevated privileges, establish persi ...[truncated 41 chars]

Remediation
View remediation

Remediation Suggestions

  • Pin the dependency to a reviewed version instead of resolving the latest available release.
  • Record dependencies in a lock file and require cryptographic hashes for downloaded artifacts.
  • Install packages only from an explicitly trusted package index.
  • Review dependency updates before changing the pinned version.
  • Use an isolated virtual environment with minimal filesystem, credential, and network access.
  • Document the separate Graphviz system-binary dependency and install it through a trusted, version-controlled system package source.
Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

Undeclared Tool Scope

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.