blog-outline

Pass

Audited by Gen Agent Trust Hub on Mar 30, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is designed for SERP-informed outline generation and follows standard data processing patterns for this use case.
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection due to its core functionality of processing untrusted web data.
  • Ingestion points: Step 2 uses WebSearch and WebFetch to collect external data from top search results.
  • Boundary markers: None present in the instructions to separate fetched data from the prompt context.
  • Capability inventory: Includes file system access (Write, Bash, Read, Grep, Glob) and web access tools (WebSearch, WebFetch).
  • Sanitization: No explicit sanitization of the retrieved web content is performed; the agent is instructed to summarize it into a structured markdown outline.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 30, 2026, 02:44 AM