Security audit

Praesidia

Security checks across malware telemetry and agentic risk

Overview

Praesidia is a coherent API integration for agent trust checks and guardrail management, with important consent and change-control precautions but no evidence of hidden or malicious behavior.

Install this only if you intend OpenClaw to use Praesidia. Use a scoped API key where possible, avoid validating secrets or regulated data unless Praesidia is authorized for that data, and explicitly review organization ID, agent ID, scope, action, severity, and rollback steps before applying guardrails.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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

Medium

Confidence: 92% confidence
Finding: The README instructs the assistant to apply guardrails that can change live agent behavior, including BLOCK, REDACT, REPLACE, RETRY, and ESCALATE, but does not require an explicit confirmation or warning before making those changes. In a skill intended to drive API calls from natural-language requests, this omission can cause unintended production-impacting policy changes, service disruption, or altered outputs if the assistant acts on ambiguous user input.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The content-validation workflow sends arbitrary user content to Praesidia's external API but does not warn that sensitive prompts, PII, credentials, or regulated data may leave the local system. Because the skill explicitly supports security and compliance use cases, users may submit highly sensitive material under the assumption it is only being checked locally, creating privacy, compliance, and data-handling risks.

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The activation description includes many broad trigger phrases such as generic security, moderation, discovery, and 'is this agent safe?' language. This can cause the skill to be invoked in contexts where the user did not intend to use an external verification service, increasing the chance of unnecessary API calls and accidental disclosure of agent identifiers or related context.

Missing User Warnings

Medium

Confidence: 97% confidence
Finding: The skill instructs sending user-provided message content to a remote '/guardrails/validate' endpoint but does not require a clear user-facing notice or consent before transmitting that content off-platform. If users submit sensitive prompts, PII, regulated data, or confidential business text for validation, the agent may exfiltrate that material to a third-party service unexpectedly.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The skill performs a state-changing POST to create/apply guardrails on a user's agent without documenting an explicit confirmation step. This can lead to unintended policy changes, service disruption, content blocking/redaction, or compliance-impacting configuration changes if the model acts on an ambiguous request or misidentifies the target agent.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.exposed_secret_literal

File appears to expose a hardcoded API secret or token.

Critical

Code: suspicious.exposed_secret_literal
Location: README.md:566