Security audit

Embeddings

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent markdown-only embeddings guide; it discusses expected third-party APIs and local vector storage, with privacy cautions users should keep in mind.

Before installing or using this skill, treat hosted embedding and reranking examples as sending the supplied text to that provider. Use local models or self-hosted storage for sensitive data, and be deliberate about where persistent vector databases are written.

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
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (5)

Missing User Warnings

Low

Confidence: 83% confidence
Finding: This markdown file includes code examples that transmit input text to external embedding services, which can affect privacy if users embed sensitive content. While the document compares providers, it does not explicitly warn readers that using the hosted APIs sends their text off-system.

Missing User Warnings

Low

Confidence: 82% confidence
Finding: The example demonstrates embedding user-provided text with a remote provider, but the surrounding markdown does not warn that the content leaves the local environment. For documentation files, safety-relevant behavior that may affect user data should be disclosed.

Missing User Warnings

Low

Confidence: 81% confidence
Finding: The markdown presents a remote embedding call but does not disclose that the text is transmitted to a third-party service. This omission matters because users may assume examples are privacy-neutral unless the documentation says otherwise.

Missing User Warnings

Low

Confidence: 85% confidence
Finding: This markdown file includes sample code that initializes a Cohere client and sends both the user's query and retrieved document contents to a third-party reranking API. The surrounding documentation does not warn that this step transmits potentially sensitive data off-system, which is a missing user warning for markdown content.

Missing User Warnings

Low

Confidence: 84% confidence
Finding: This markdown file includes a Chroma example using `PersistentClient(path="./chroma_db")`, which would store embeddings and related data on local disk. The surrounding description does not warn users that this behavior may persist potentially sensitive document data, which is the kind of data-affecting behavior SQP-2 covers for markdown files.

Static analysis

No suspicious patterns detected.