Security audit

simulated-peak-plot

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly matches a peak-plot generator, but its script secretly alters generated data and PNG pixels in ways users are not told about.

Review before installing. The skill does not show network access or credential theft, and normal CSV/PNG/CSV file operations fit its purpose, but generated outputs may contain undisclosed deterministic markers. Use only if you are comfortable with that behavior, or remove the fixed-seed noise replacement and PNG pixel post-processing first.

SkillSpector

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
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 (4)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 85% confidence
Finding: The skill declares low sensitivity and no meaningful permissions, yet its documented behavior includes reading CSV input and writing PNG/CSV outputs. This mismatch can mislead reviewers and users about the actual file-system access the skill requires, weakening permission governance and making unexpected data access harder to audit.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 98% confidence
Finding: The detected hidden behaviors—embedding bits in PNG pixel LSBs, placing a seeded noise sequence in the blue channel, and replacing the first 300 output samples with reproducible noise—are unrelated to legitimate peak-plot generation. These are classic steganographic and covert-signaling patterns that can hide markers, encode data, or watermark outputs without user consent, creating a strong risk of covert data exfiltration or deceptive output manipulation.

Context-Inappropriate Capability

High

Confidence: 99% confidence
Finding: The script injects a deterministic pseudorandom sequence into the first 300 signal samples using a fixed seed, which is unrelated to legitimate peak simulation. This creates a covert watermark/fingerprint in exported data that can track outputs or carry hidden information without user awareness, making the behavior deceptive and unsafe in a plotting skill.

Context-Inappropriate Capability

High

Confidence: 99% confidence
Finding: After saving the PNG, the code reopens it and alters specific pixels to encode bit patterns and image-domain noise derived from a fixed seed. This is classic hidden-data embedding unrelated to rendering a plot and can be used for covert signaling, watermarking, or output fingerprinting without consent.

Static analysis

No suspicious patterns detected.