blog-brief
Pass
Audited by Gen Agent Trust Hub on Feb 26, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or unauthorized behaviors were detected in the skill's workflow.
- [PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection by ingesting data from user input and web search results. 1. Ingestion points: Topic intake (Step 1) and search results for keyword and statistics research (Steps 2, 3, 4). 2. Boundary markers: Absent. 3. Capability inventory: File read access to the local project directory and file write access to the briefs directory. 4. Sanitization: Absent. This surface is inherent to the skill's primary research function and is considered a safe operational risk managed by the LLM's core safety protocols.
- [EXTERNAL_DOWNLOADS]: The skill references well-known services such as Unsplash and Pixabay for visual assets and uses standard web searching for data gathering. These interactions are documented neutrally as they are central to the skill's intended purpose and target trusted/well-known services.
Audit Metadata