dbs-title-cover-intro
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user-provided content such as manuscripts, transcripts, and drafts to generate social media assets.
- Ingestion points: User input provided during the conversation is ingested for analysis and rewriting across all core modules (
SKILL.md,references/material-selection.md). - Boundary markers: The instructions do not define explicit structural delimiters (e.g., XML tags or unique separators) between instructions and user-provided data.
- Capability inventory: The skill is limited to text generation and logical reasoning. It does not utilize tools for file system access, network operations, or shell execution.
- Sanitization: The skill includes a dedicated validation reference (
references/promise-check.md) that enforces factual accuracy and checks that the generated titles fulfill the promises made in the content. - [EXTERNAL_DOWNLOADS]: The skill references academic papers and theoretical documentation from reputable educational and publishing institutions.
- Evidence: Links to PDF documents hosted on trusted domains including Sage Publishing (
sagepub.com), the University of North Carolina (unc.edu), and George Mason University (gmu.edu) are provided inreferences/angle-theory-sources.mdto support the theoretical framework used for content analysis.
Audit Metadata