copy-editing
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data, creating a potential surface for indirect prompt injection attacks. \n
- Ingestion points: Data enters the agent context through the
sandbase_runtool, which is used for "external evidence, search, scraping, [and] enrichment" as specified in the SandBase boundary section of SKILL.md. \n - Boundary markers: The skill instructions do not specify the use of delimiters, XML tags, or protective prompts to isolate retrieved external content from the agent's core instructions. \n
- Capability inventory: The skill has the ability to read local files, execute vendor-specific capabilities via
sandbase_run, and generate deliverables. \n - Sanitization: There are no explicit requirements for the agent to sanitize, filter, or escape data fetched from external sources before using it in the copy-editing workflow.
Audit Metadata