Back to skill

Security audit

Long Task Handoff

Security checks across malware telemetry and agentic risk

Overview

This skill persistently writes local handoff notes for long-running agent work, which matches its stated purpose and is disclosed.

Install this only if you want agents to create persistent workspace handoff files during long tasks or context compaction. Review generated handoffs before committing or sharing a repository, because they may contain task context, file paths, branch state, and summaries of user instructions even though the skill tries to exclude secrets.

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
82% confidence
Finding
The eval at line 36 uses a very broad trigger phrase, effectively treating a generic user request like 'continue yesterday's ... task' plus the presence of ACTIVE.md as sufficient to resume prior context. In a security-sensitive agent, this can cause unintended activation of handoff recovery behavior, potentially importing stale, attacker-planted, or unrelated workspace state without strong confirmation that the handoff is fresh and belongs to the current user task.

Missing User Warnings

Low
Confidence
71% confidence
Finding
The script creates or overwrites handoff files and updates `handoffs/ACTIVE.md` automatically with no user-facing disclosure or confirmation. In an agent setting, silent persistence can expose task context, file paths, branch state, and user-provided content to later sessions or other local readers, which is more sensitive than an ordinary temp file because the feature is designed to carry forward conversational state.

VirusTotal

37/37 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.