Security audit

gladia-sdk-integration

Security checks for vulnerabilities and agentic risk

Overview

This skill is documentation for installing and using Gladia SDKs, with no hidden execution or persistence behavior found.

Install only if you intend to build with Gladia's speech-to-text SDKs. Treat API keys and audio as sensitive: keep keys server-side when possible, get consent before recording or streaming microphone audio, and confirm before calling delete methods because they remove stored job or session data.

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

Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
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 (7)

Whitespace Padding

Medium

Category: Prompt Injection
Confidence: 70% confidence
Finding: Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.
Content: Scanner excerpt · SKILL.md (reported line 26)May include surrounding context.
md
## When to Use SDK vs Raw API | Scenario | Approach | | ---------------------------------------- | ------------------------------------------------------------------------------------------------------------- | | Any JS/TS or Python project | **SDK** — always | | Browser app | **SDK** — JS SDK supports ESM/IIFE bundles |

Whitespace Padding

Medium

Category: Prompt Injection
Confidence: 70% confidence
Finding: Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.
Content: Scanner excerpt · SKILL.md (reported line 28)May include surrounding context.
md
| Scenario | Approach | | ---------------------------------------- | ------------------------------------------------------------------------------------------------------------- | | Any JS/TS or Python project | **SDK** — always | | Browser app | **SDK** — JS SDK supports ESM/IIFE bundles | | Need custom HTTP client or middleware | SDK first; use `httpHeaders` / `httpTimeout` config. Fall back to raw REST only if SDK config is insufficient | | Language without an SDK (Go, Java, etc.) | Raw REST/WebSocket (SDK unavailable) |

Whitespace Padding

Medium

Category: Prompt Injection
Confidence: 70% confidence
Finding: Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.
Content: Scanner excerpt · SKILL.md (reported line 32)May include surrounding context.
md
| Browser app | **SDK** — JS SDK supports ESM/IIFE bundles | | Need custom HTTP client or middleware | SDK first; use `httpHeaders` / `httpTimeout` config. Fall back to raw REST only if SDK config is insufficient | | Language without an SDK (Go, Java, etc.) | Raw REST/WebSocket (SDK unavailable) | | User explicitly requests raw calls | Raw REST/WebSocket | | CI script or one-off curl test | Raw REST is acceptable | | Terminal one-off with gladia-cli on PATH | **CLI** — see [gladia-using-cli](../gladia-using-cli/SKILL.md) |

Whitespace Padding

Medium

Category: Prompt Injection
Confidence: 70% confidence
Finding: Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.
Content: Scanner excerpt · SKILL.md (reported line 33)May include surrounding context.
md
| Need custom HTTP client or middleware | SDK first; use `httpHeaders` / `httpTimeout` config. Fall back to raw REST only if SDK config is insufficient | | Language without an SDK (Go, Java, etc.) | Raw REST/WebSocket (SDK unavailable) | | User explicitly requests raw calls | Raw REST/WebSocket | | CI script or one-off curl test | Raw REST is acceptable | | Terminal one-off with gladia-cli on PATH | **CLI** — see [gladia-using-cli](../gladia-using-cli/SKILL.md) | When in doubt, use the SDK.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 92% confidence
Finding: The skill documents destructive delete(jobId) and delete(sessionId) operations, but does not clearly warn that these actions permanently remove stored transcription/session data. In an agent-skill context, omission of a deletion warning can cause unintended data loss if an assistant suggests or invokes cleanup commands without making the consequence explicit.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 91% confidence
Finding: The Node microphone streaming example demonstrates capturing live audio and sending it to a remote transcription service without any accompanying notice about consent, privacy, or legal obligations. In documentation, this can normalize deployment patterns that record speakers without adequate disclosure, increasing the risk of unauthorized collection and transmission of sensitive audio.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 92% confidence
Finding: The microphone streaming example captures live audio and immediately transmits it with session.send_audio(data) without any user-facing notice, consent prompt, or clear indication that recording/network transmission is occurring. In SDK documentation, this can lead downstream developers to copy an implementation that silently streams microphone data, creating privacy and compliance risks for end users.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.