Back to skill

Security audit

SafeHarbor Check

Security checks for vulnerabilities and agentic risk

Overview

The skill is rated benign with high confidence: the scanners found one install-command signature in its documentation, while the verdict treats its behavior as purpose-aligned and disclosed.

The verdict is benign, but benign does not mean risk-free. Before installing, consider that setup uses pip to install from a GitHub repository; this scan does not independently establish the safety of that repository or its dependencies.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • 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 (1)

YARA rule 'agent_skill_remote_bootstrap_execution': Remote script or code download followed by execution/bootstrap installation [agent_skills]

High
Category
YARA Match
Confidence
85% confidence
Finding

YARA rule matched a known malware signature (reverse shell, backdoor, ransomware, C2 framework, or info stealer).

Content

Scanner excerpt · SKILL.md (reported line 23)May include surrounding context.

classes, (A)–(R), in 45 CFR 164.514(b)(2). check reports per-class results: violations it can detect mechanically, and — in the same report, every time — what cannot be verified mechanically, marked CANNOT VERIFY and never reported as passing. apply performs the rule's three mechanical fixes on columns you declare, and nothing else. A checker, not a certifying tool.

Install

bash
pip install git+https://github.com/RakTikoo/safeharbor-check

This provides the safeharbor-check command.

Run

bash
safeharbor-check check patients.csv
safeharbor-check apply patients.csv --out cleaned.csv --dates admission_date --zips zip_code --ages age --drop patient_name,phone_number

Both commands accept --delimiter CHAR (comma by default). Use --delimiter "\t" or an actual tab for TSV. The declared character is used for input and apply output; it is never guessed. Other values must be one character, excluding CR, LF and NUL, or the command returns a one-line error