daily-agent

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed always-on dispatcher for other work modes, with no executable code or credential/network behavior, but users should understand it can automatically change the agent’s behavior.

Install this only if you want a default behavior layer that automatically switches the agent into different work modes. Review any referenced child skills separately, because this package names local child-skill paths but does not include their instructions.

SkillSpector

By NVIDIA

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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

Findings (2)

Vague Triggers

High

Confidence: 97% confidence
Finding: The skill is configured as a default, always-on dispatcher that auto-selects behaviors based on broad conversational context rather than an explicit user invocation. This creates prompt-boundary ambiguity and increases the chance that unrelated user text, embedded content, or adversarial instructions inside analyzed materials can silently change the agent’s operating mode and bypass user expectations.

Vague Triggers

Medium

Confidence: 94% confidence
Finding: Several child-skill triggers are subjective and expansive, such as inferred impatience, general bug descriptions, or the agent’s own judgment that a mode is suitable. Broad heuristics like these can be misfired by normal conversation or attacker-crafted text, causing unintended behavior changes and making the agent easier to steer indirectly.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal