Security audit

gladia-audio-intelligence

Security checks for vulnerabilities and agentic risk

Overview

This is a documentation-only skill for configuring Gladia audio intelligence features, with privacy considerations users should manage before sending sensitive audio or transcripts.

Before using this skill on real recordings, confirm that sending audio, transcripts, metadata, and audio-to-LLM prompts to Gladia or downstream model processing is allowed by your policy and by participant consent. Minimize submitted data, enable PII redaction when appropriate, and avoid sensitive metadata or regulated content unless retention and compliance requirements are clear.

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

Medium

Category: Not specified by scanner
Confidence: 90% confidence
Finding: The skill documents features such as translation, NER, summarization, and audio-to-LLM without warning that user audio and derived transcripts may contain sensitive or regulated data and will be sent to an external service for processing. This can lead developers to enable these features on confidential recordings without informed consent, data-minimization checks, or compliance review, increasing privacy and data-handling risk.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 92% confidence
Finding: This markdown file documents starting a live session with translation and other real-time analysis features, which necessarily involve sending audio/transcript content to a remote service for processing. The page provides configuration details but does not warn users that enabling these features may affect privacy or involve transmission of potentially sensitive spoken data.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 89% confidence
Finding: These features derive or attach potentially sensitive information from live conversations, including sentiment, named entities, summaries, and arbitrary metadata. The documentation explains how to enable them but omits any caution about privacy, consent, or the risk of storing or exposing personal or contextual data.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 91% confidence
Finding: This documentation encourages use of PII redaction on sensitive audio/transcripts but does not warn that the source data may still contain regulated personal data before redaction, may be transmitted to a third-party service, and may require consent, retention controls, and policy review. In a skill focused on audio intelligence, omission of privacy guidance can lead users to process sensitive recordings under the false assumption that enabling redaction alone fully addresses privacy and compliance risk.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 94% confidence
Finding: The Audio-to-LLM section instructs users to run custom prompts against transcripts without warning that transcript contents may be exposed to downstream model processing, potentially including sensitive or regulated data. In this skill context, that is more dangerous because users are explicitly encouraged to submit arbitrary conversation content to an LLM feature, which can create confidentiality, compliance, and data-handling risks if users assume it is safe by default.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.