copy-editing
Pass
Audited by Gen Agent Trust Hub on Mar 13, 2026
Risk Level: SAFENO_CODEPROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or dangerous command executions were detected. The skill primarily consists of natural language instructions for text editing and a static reference file.
- [PROMPT_INJECTION]: The skill processes untrusted user input (marketing copy), which represents a surface for indirect prompt injection.
- Ingestion points: User-provided text for editing and the local context file
.claude/product-marketing-context.md. - Boundary markers: No explicit delimiters are used to separate untrusted copy from instructions.
- Capability inventory: The skill is restricted to file read operations; no network access, subprocess execution, or file writing capabilities are present.
- Sanitization: No sanitization is performed on the input text.
Audit Metadata