research-methodology

Pass

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: LOWNO_CODE
Full Analysis
  • [SAFE] (SAFE): Analysis of all skill files confirms they are instructional Markdown. No scripts, binaries, or hidden malicious patterns were detected.
  • [PROMPT_INJECTION] (LOW): This skill defines a workflow for processing untrusted external content (web research), which is an inherent risk of the task. 1. Ingestion points: WebSearch and WebFetch results. 2. Boundary markers: Absent. 3. Capability inventory: Local file-write to project references. 4. Sanitization: Absent. This surface is managed by the skill's emphasis on official sources, but the agent remains susceptible to indirect injection if source content contains malicious instructions.
Audit Metadata
Risk Level
LOW
Analyzed
Feb 16, 2026, 04:18 AM