Security audit

Signal Detector

Security checks for vulnerabilities and agentic risk

Overview

This note-taking skill is coherent, but it would automatically save exact wording from most user messages into persistent notes, so it needs review before installation.

Install only if you intentionally want ambient knowledge capture across conversations. Before using it with sensitive work, confirm where notes are stored, who can access them, how to disable capture, how to delete captured entries, and whether exact quotes can be replaced with summaries or redacted by default.

A.I.G

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

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Missing User Warnings

High

Confidence: 97% confidence
Finding: The skill is explicitly designed for always-on capture of inbound user messages and describes logging/capture behavior without any user-facing notice, consent, or privacy boundary. Because it operates ambiently and persistently, users may disclose sensitive personal, business, or credential-adjacent information that is then retained unexpectedly.

Ssd 3

High

Confidence: 99% confidence
Finding: The skill mandates collection from every substantive user message, in parallel, with exact phrasing and provenance, which amounts to broad surveillance-style harvesting of natural-language content. In context, this is more dangerous because it is silent, continuous, and framed as non-blocking background behavior, making overcollection and unnoticed retention very likely.

Missing User Warnings

High

Confidence: 98% confidence
Finding: These instructions direct the agent to write user-derived content into notes, project indexes, dashboards, or the knowledge base, creating persistent storage of user messages and inferred ideas. Persisting exact user phrasing and cross-linked facts without warning or consent materially increases privacy, confidentiality, and downstream disclosure risk.

Ssd 3

High

Confidence: 98% confidence
Finding: The required output format logs exact quoted user phrasing, encouraging verbatim retention of potentially sensitive statements, secrets, regulated data, or proprietary ideas. Quoted logging also increases the chance that confidential content is later surfaced through search, sharing, sync, backups, or other downstream knowledge-base operations.

Autonomous Decision Making

Medium

Category: Excessive Agency
Content: - Paraphrasing or smoothing the user's language — keep the original voice - Blocking the main conversation to finish capture — spawn or defer - Creating knowledge pages for every passing mention — apply the notability gate - Over-writing existing knowledge base material without checking what's there first - Capturing purely operational exchanges as signal
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.

Static analysis

No suspicious patterns detected.