ClawHub
Back to skill

Security audit

Session Health Monitor

Security checks across malware telemetry and agentic risk

Overview

The skill is mostly local and related to context monitoring, but it includes an under-documented setup script that changes Claude settings to run a statusline script in future sessions.

Install only if you are comfortable with local session facts being written to memory files. Review any facts before snapshotting secrets, set MEMORY_DIR deliberately, and do not run scripts/setup-statusline.sh unless you accept changes to ~/.claude/settings.local.json and know how to revert from the backup it creates.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Rogue AgentSelf-Modification, Session Persistence
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Tp4

High
Category
MCP Tool Poisoning
Confidence
84% confidence
Finding
The skill's stated behavior does not fully match the described implementation: it claims Telegram threshold warnings, but the file only instructs agents to append footers and references external scripts/tools that are not shown here. More importantly, the described behavior includes modifying local config/state outside the advertised scope, which can surprise operators and weaken trust boundaries around agent skills.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The installer writes to ~/.claude/settings.local.json and changes the statusLine command without any interactive confirmation or dry-run step. Although this is framed as a convenience installer, silently modifying a user's agent configuration can unexpectedly alter behavior and create persistence for a script execution path under the user's account.

Session Persistence

Medium
Category
Rogue Agent
Content
session_id="${session_id:-unknown}"
state_file="/tmp/session-health-${session_id}.json"

# Load previous state
prev_pct=0
compactions=0
if [[ -f "$state_file" ]]; then
Confidence
94% confidence
Finding
Load previous state

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.