ai-product-teardown
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's research-first workflow involves ingesting untrusted data from the web, such as user reviews and blog posts, to inform the teardown. While this creates an attack surface for indirect prompt injection, the skill lacks high-privilege capabilities (e.g., file system writes, command execution) that could be exploited by malicious content found online.
- Ingestion points: SKILL.md (Research-First Workflow section) specifies searching for latest product updates, user reviews, and blog posts.
- Boundary markers: Absent. The instructions do not specify using delimiters or ignore-instructions warnings when processing search results.
- Capability inventory: No subprocess calls, file-write operations, or dangerous network operations detected in the skill instructions.
- Sanitization: Absent.
- [EXTERNAL_DOWNLOADS]: The skill is designed to fetch and cite external data via web searches to perform its analysis. This is a core part of its intended functionality and does not involve downloading executable code or scripts.
Audit Metadata