outreach-assemble
Pass
Audited by Gen Agent Trust Hub on Aug 4, 2026
Risk Level: SAFECREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns detected. The skill follows standard practices for the Paperclip platform ecosystem.
- [CREDENTIALS_UNSAFE]: The skill documentation describes how the platform's tools interact with configuration files like ~/.paperclip/auth.json and ~/.paperclip/context.json to manage authentication. These are standard locations for the platform tools and the skill includes instructions for secure token handling.
- [EXTERNAL_DOWNLOADS]: The skill documentation instructs the setup of the paperclipai CLI via npm install and the use of the @bbengamin/paperclip-mcp-server package via npx. These are vendor-provided resources associated with the skill's author.
- [COMMAND_EXECUTION]: The personalization workflow involves a driver script, scripts/ralph-run.sh, as defined in the platform's run contract.
- [PROMPT_INJECTION]: The skill processes lead data enriched from external sources for message personalization, presenting a surface for indirect prompt injection. 1. Ingestion points: Enriched lead data fields in the Twenty CRM. 2. Boundary markers: None explicitly defined. 3. Capability inventory: Drafting campaign assets and writing CRM references. 4. Sanitization: None mentioned.
Audit Metadata