Triskill

Security checks across malware telemetry and agentic risk

Overview

Triskill is a disclosed utility skill for fact-checking, bounded command retry diagnostics, and local agent coordination, with some local logging and storage users should understand.

Install only if you want these three utilities available to your agent. Do not use the self-heal wrapper for commands that may print secrets or perform sensitive side effects, and do not store credentials in shared memory because the local files are not encrypted.

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

Medium

Confidence: 83% confidence
Finding: The trigger language is broad and includes common phrases like fact-checking, failure recovery, and multi-agent coordination, which can cause the skill to be invoked in many routine contexts. Because the skill can reach the network, run user-supplied commands, and write local state, over-broad invocation meaningfully raises the risk of accidental exposure to higher-risk operations.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The wrapper writes command strings and tails of stdout/stderr to selfheal_log.jsonl without any runtime warning or consent mechanism. Since commands and output often contain secrets, tokens, filesystem paths, or proprietary data, this can create an inadvertent local data-exposure trail, especially in multi-agent/shared-workspace use where other sessions may read the log.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal