Back to skill

Security audit

Music Toolkit Free

Security checks for vulnerabilities and agentic risk

Overview

This skill is a local music/MIDI/audio helper with expected file-writing and install guidance, and I found no hidden persistence, exfiltration, or destructive behavior.

Install this only if you are comfortable letting an agent run local Python/audio tooling and write generated media files. Use a virtual environment, pin package versions where possible, install FFmpeg from trusted sources, and choose explicit output paths so generated files do not overwrite existing work.

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

Warning
Location
SKILL.md:257
Finding

Unpinned Third-Party Python Dependencies

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 257–259 and 437–439
Vulnerability Type: Supply-chain risk from unpinned third-party dependencies
Risk Level: Medium

Vulnerable code at lines 257–259:

bash
# Python music processing libraries
pip install midiutil pydub numpy

Repeated installation guidance at lines 437–439:

text
| midiutil | Python library | MIDI generation | `pip install midiutil` |
| pydub | Python library | Audio processing | `pip install pydub` |
| numpy | Python library | Signal processing | `pip install numpy` |

Technical Analysis

The installation instructions do not pin reviewed package versions, validate package hashes, use a lockfile, or require a trusted package index. Consequently, package resolution depends on the mutable state of the Python package index and the user's local pip configuration at installation time.

If an upstream package, dependency, or configured package index is compromised, an attacker-controlled release could be selected. Malicious package code may execute during installation through build hooks or later when the installed module is imported. The absence of integrity hashes prevents the installer from detecting that the downloaded artifacts differ from previously reviewed artifacts.

No malicious dependency, hostile index, or active compromise was observed in the audited project. The finding concerns the unsafe dependency acquisition process documented by the skill.

Attack Path

  1. A user or AI agent follows the instructions and runs pip install midiutil pydub numpy.
  2. pip resolves package versions and transitive dependencies using its currently configured index.
  3. An attacker compromises an upstream release, a transitive dependency, or an index available through the local pip configuration.
  4. The resolver downloads the attacker-controlled artifact because no exact version or cryptographic hash is require ...[truncated 838 chars]
Remediation
View remediation

Remediation Suggestions

  1. Move dependencies into a committed requirements or lock file and pin every direct and transitive dependency to a reviewed version.
  2. Record cryptographic hashes for all approved distribution artifacts and enforce them:
    bash
    python -m pip install --require-hashes -r requirements.txt
    
  3. Generate the locked dependency set with a reproducible dependency-management tool and review changes before updating it.
  4. Explicitly document and enforce an approved HTTPS package index rather than relying on arbitrary local pip configuration.
  5. Install dependencies inside an isolated virtual environment or container running with minimal filesystem and network privileges.
  6. Scan locked dependencies for known vulnerabilities and periodically update pins after security review and compatibility testing.
  7. Replace every repeated unpinned command in SKILL.md with the hardened installation procedure so users are not directed to bypass the lock file.
Vulnerability Patterns
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (7)

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill is presented as a local music generation/editing toolkit, but it requests the powerful exec capability and explicitly classifies itself as MD+EXEC. That creates a capability mismatch: natural-language instructions for benign music tasks can be expanded into arbitrary shell execution, increasing the risk of command injection, unintended system changes, or abuse beyond the stated scope.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The documentation says the free version does not support AI arranging, yet the dependency table states an LLM API is required. This inconsistency can mislead users and reviewers about whether prompts, file contents, or metadata may be sent to a model-backed service, creating hidden data exposure and trust-boundary issues.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill advertises export/save/convert operations but does not warn that these actions create or modify local files. In a tool with write and exec access, silent file output can cause overwrites, clutter, or unexpected modification of user data, especially when paths or filenames are inferred automatically.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The examples directly write files such as melody.mid and output.mp3 without any safety guidance about file creation, path selection, or overwrite behavior. Users or agents may copy these patterns into automated workflows, leading to unintended writes in the current working directory or replacement of existing files.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger conditions are broad enough to activate the skill for generic video/audio editing and media conversion requests, not just narrowly defined music tasks. In a skill that also has write and exec privileges, over-broad activation raises the chance of unintended invocation and execution on user files outside the expected context.

Content

No source excerpt is available for this finding.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
70% confidence
Finding

Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Content

Scanner excerpt · SKILL.md (reported line 263)May include surrounding context.

...

系统依赖 (FFmpeg)

macOS: brew install ffmpeg

Ubuntu: sudo apt install ffmpeg

Windows: 下载 FFmpeg 并配置 PATH

text

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The overview states that all processing is completed locally, and the API key section reiterates that no external API key is needed because processing is local. However, the error handling section includes a 网络错误 case with instructions to test connectivity, proxies, and firewalls, which conflicts with the claimed fully local-only processing model.

Content

No source excerpt is available for this finding.