investigate-cve
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external web sources and internal database fields, creating a potential surface for indirect prompt injection. 1. Ingestion points: CVE advisories from WebSearch and WebFetch, and CVE metadata descriptions from subimageRunCypher. 2. Boundary markers: Not explicitly specified in the report structure. 3. Capability inventory: Ability to query the environment graph via subimageRunCypher and perform web searches. 4. Sanitization: The skill provides clear instructions to escape single quotes and backslashes for user-provided identifiers.
- [DYNAMIC_EXECUTION]: The skill relies on the dynamic generation of database queries. Evidence: SKILL.md contains several template strings for Cypher queries where user input is inlined. The author includes specific security guidance to mitigate query injection by sanitizing inputs before execution.
Audit Metadata