copywriting
Pass
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: LOW
Full Analysis
- [Prompt Injection] (SAFE): No malicious instructions, jailbreak patterns, or system prompt extraction attempts were detected. The use of 'IMPORTANT' in documentation is contextually appropriate for user guidance.
- [Data Exposure & Exfiltration] (SAFE): The skill only references a local context file (
.claude/product-marketing-context.md) which is standard for AI workspace environments. No sensitive file paths (e.g., SSH keys, AWS creds) or network calls were found. - [Indirect Prompt Injection] (INFO): The skill possesses an ingestion surface by processing user-provided copy and local context. However, because it lacks side-effect capabilities (no command execution, network access, or file modification), the risk is limited to influencing the agent's reasoning within the current session.
- [Remote Code Execution] (SAFE): No package installations (pip, npm) or remote script downloads (curl | bash) are present. The skill consists entirely of Markdown guidance.
- [Obfuscation] (SAFE): Content is clear and human-readable with no evidence of Base64 encoding, zero-width characters, or homoglyph attacks.
Audit Metadata