copy-editing
Pass
Audited by Gen Agent Trust Hub on Jun 21, 2026
Risk Level: SAFE
Full Analysis
- [DATA_EXFILTRATION]: Reads from the local file
.claude/product-marketing-context.mdto guide editing choices. This access is restricted to the local environment for context-aware editing and does not involve any external network transmission or data exfiltration. - [PROMPT_INJECTION]: The skill ingests untrusted text provided by users for editing (Ingestion Point). While it lacks explicit boundary markers or sanitization, the skill possesses no capabilities for code execution, network requests, or file writes (Capability Inventory), which prevents potential indirect prompt injections from causing harm to the system or data.
Audit Metadata