Pilot Scientific Research Team Setup

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:42 PM UTC
• Version:
  1.0.0

Overview

This is an instruction-only setup skill for a multi-agent research workflow, with external report publishing disclosed and aligned to the report-writer role.

Install this only if you trust the `pilotctl` and `clawhub` tools and the listed pilot-* skills. Before enabling the report role or webhook bridge, decide what research data may leave your environment, restrict destinations where possible, and manually review reports before publishing externally.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (4)

Context-Inappropriate Capability

Medium

• Confidence: 90% confidence
• Finding: The report role includes outbound publication and webhook capabilities that materially expand the trust boundary from internal collaboration to external data egress. In a research pipeline, this can expose unpublished results, sensitive datasets, or derived findings if the report agent is misconfigured or triggered before review.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The README normalizes publishing reports externally over port 443 without warning about confidentiality, approval, or data-governance controls. Because scientific workflows often handle unpublished, proprietary, or regulated material, omitting disclosure safeguards increases the risk of accidental external release.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The example workflow culminates in publishing a final report externally and presents this as a normal success path without cautionary language or review checkpoints. Examples strongly influence operator behavior, so this can lead users to expose internal findings or sensitive research artifacts unintentionally.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The skill configures a reporting agent that can send data externally via pilot-webhook-bridge and publish reports over port 443, but the top-level skill description does not warn users that research outputs may be transmitted outside the local agent mesh. In a research workflow, unpublished results, proprietary data, or sensitive findings could be unintentionally disclosed if the user enables this setup without clear notice.

VirusTotal

66/66 vendors flagged this skill as clean.