knowledge-synthesis
Pass
Audited by Gen Agent Trust Hub on Apr 7, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill consists of natural language instructions for data synthesis and does not contain any executable code, sensitive file access, or network operations. It follows standard patterns for instruction-based agent skills.
- [PROMPT_INJECTION]: The skill is designed to process untrusted data from external sources like chat, email, and documents, which is a potential surface for indirect prompt injection.
- Ingestion points: Processes results from external search tools as described in
SKILL.md. - Boundary markers: The skill uses source-specific markers (e.g.,
~~chat,~~email) to identify data origins, which provides a basic level of context separation. - Capability inventory: No sensitive capabilities such as file system writes, network access, or command execution are requested or used by this skill.
- Sanitization: The instructions focus on narrative synthesis and deduplication; while they do not explicitly mention sanitization of embedded instructions, the lack of dangerous tools minimizes the risk of exploitation.
Audit Metadata