Sentinel - AI Agent State Guardian

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:18 PM UTC
• Version:
  1.0.1

Overview

Sentinel is a local backup and integrity-monitoring skill whose file access and restore behavior match its stated purpose, but users should configure automatic restore carefully.

Install only if you want a local tool that reads, hashes, copies, and may restore selected workspace files. Set WORKSPACE_ROOT and CRITICAL_FILES narrowly, store backups in a protected location, and disable or avoid automatic restore until you trust the baseline because restore operations can overwrite current agent memory, config, or state.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (6)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 94% confidence
• Finding: The skill clearly instructs the agent to read, hash, back up, restore, and quarantine workspace files, which implies file read/write capability, but the metadata does not declare those permissions. Hidden or undeclared file access is dangerous because users and hosting systems cannot accurately assess or constrain what the skill may do before use.

Missing User Warnings

Medium

• Confidence: 87% confidence
• Finding: The README promotes automatic self-healing and restoration of files, which means the tool can modify workspace data without strong upfront warnings about overwrite, rollback, or false-positive recovery risks. In an AI agent workspace, unexpected restoration can revert legitimate state, destroy recent updates, or mask tampering by replacing current files with older copies.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: The skill advertises automatic restore and quarantine behavior as protective features but does not clearly warn that these actions can overwrite current files or move user data without interactive confirmation. In a workspace-management skill, that omission increases the risk of unintended data loss, rollback of legitimate changes, or disruption of active agent state.

Missing User Warnings

Low

• Confidence: 82% confidence
• Finding: The backup feature copies workspace contents to another location, but the documentation does not clearly warn that sensitive data may be duplicated outside the primary workspace. This can expand the exposure surface for secrets, memory files, or agent state if the backup directory has weaker protections or different retention rules.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The integrity violation response options include auto-restore and quarantine, both of which can change or relocate files automatically, yet the section does not clearly frame them as potentially destructive operations. Because these actions may trigger in response to false positives or stale state, the skill could unexpectedly revert legitimate work or move files out of place.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The auto-restore path unconditionally overwrites the target file with the latest backup via shutil.copy2(backup_path, filepath) after only limited checks, with no user confirmation, no dry-run mode, and no validation that the selected backup is the intended safe version. In an agent workspace, this can silently replace current state or legitimate updates, causing integrity loss, rollback to stale data, or destructive behavior if restoration is triggered incorrectly or by manipulated monitoring inputs.

VirusTotal

66/66 vendors flagged this skill as clean.