Error Recovery

Security checks across malware telemetry and agentic risk

Overview

This is a coherent error-recovery skill, but users should limit automatic credential handling and sanitize any saved error logs.

Install if you want structured error handling and bounded retry guidance. Configure the agent to ask before using tokens, changing authentication, running privileged fixes, updating workflow files, or writing persistent logs, and redact secrets from any stored error records.

SkillSpector

By NVIDIA

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration

Findings (5)

Context-Inappropriate Capability

Medium

Confidence: 92% confidence
Finding: The skill instructs the agent to create persistent error logs containing context, causes, and fixes, and elsewhere suggests updating workflow artifacts. For an error-recovery skill, this expands behavior from transient recovery guidance into durable state changes and process modification, which can create unauthorized data retention and scope creep beyond the user’s request.

Intent-Code Divergence

Medium

Confidence: 88% confidence
Finding: The skill correctly says authentication errors should not be auto-retried, but its example then proposes trying token authentication automatically. That inconsistency can normalize autonomous credential-handling behavior, leading agents to switch auth methods or use available secrets without explicit user consent.

Vague Triggers

High

Confidence: 95% confidence
Finding: The frontmatter trigger list includes very common words such as 'error', 'failed', and 'retry', which are likely to appear in normal conversation, logs, and unrelated tasks. This makes the skill prone to over-activation, causing it to override normal agent behavior and potentially invoke recovery actions in contexts where they are not appropriate.

Vague Triggers

High

Confidence: 95% confidence
Finding: The listed trigger phrases are broad, multilingual, and unconstrained, so the skill could fire on ordinary text mentioning errors rather than actual runtime failures. In an agent setting, ambiguous triggering increases the chance of unrequested retries, reporting behaviors, or state changes based on incidental language.

Ssd 3

Medium

Confidence: 97% confidence
Finding: The skill tells the agent to include full error messages, what it was doing, and to persist detailed error logs. Error output and operational context often contain secrets, personal data, file paths, tokens, prompts, or internal system details, so this guidance can amplify sensitive-data exposure both in user-facing responses and in stored logs.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal