MedPilot

Security checks across malware telemetry and agentic risk

Overview

MedPilot is a coherent local single-patient health-tracking skill, but it handles sensitive medical data and assumes use of a trusted local MedPilot CLI/API.

This skill appears purpose-aligned and locally scoped. Before using it, make sure the local MedPilot CLI/API is trusted, keep any MedPilot token private, and confirm medication or health-record writes before the agent runs them.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Risk analysis

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

#
ASI06: Memory and Context Poisoning
Medium
What this means

Your medication orders, intake history, and health readings may be entered into local records and reports.

Why it was flagged

The skill asks the agent to collect and use private health details for tracking, reminders, alerts, and reports, so any stored or reused MedPilot state should be treated as sensitive.

Skill content
- patient identity for this local instance
- doctor order text
- medication intake updates
- home metrics such as blood pressure or glucose
Recommendation

Use the skill only with a trusted local deployment, understand where local data is stored, and avoid entering unrelated medical details.

#
ASI02: Tool Misuse and Exploitation
Medium
What this means

A mistaken command could record the wrong medication, time, or patient state.

Why it was flagged

The quickstart documents API calls that can write medication intake records. This is aligned with the skill purpose, but incorrect parameters could create inaccurate health logs.

Skill content
curl -X POST http://localhost:3456/api/intakes   -H 'Content-Type: application/json'   -H 'X-MedPilot-Token: <token>'
Recommendation

Have the agent summarize patient IDs, medication IDs, and record changes before writing them, especially for intake or skipped-dose logs.

#
ASI03: Identity and Privilege Abuse
Low
What this means

If the token is exposed or used against the wrong local instance, someone or something could access or change health records.

Why it was flagged

The local API examples use an authorization token. That is expected for an API workflow, but the token grants access to the local MedPilot instance and should be handled as a credential.

Skill content
-H 'X-MedPilot-Token: <token>'
Recommendation

Keep the MedPilot token private, provide it only when needed, and rotate or revoke it if it is pasted into an untrusted place.

#
ASI04: Agentic Supply Chain Vulnerabilities
Low
What this means

Running the command from an untrusted directory could execute unrelated npm scripts or operate on an unintended local project.

Why it was flagged

The skill package is instruction-only but its examples rely on a separate local npm project or CLI. The safety of that external local project is not established by these artifacts.

Skill content
npm run dev -- ingest-order --patient <patient-id> --text
Recommendation

Run the CLI examples only inside a trusted MedPilot project and review the local package scripts before use.