cold-email-copywriting
Fail
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: CRITICALINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted user data in Phase 1 (Intake) to generate personalized email copy. This data includes target persona descriptions, pain points, and specific trigger signals. While the skill provides structured output templates, it lacks explicit instruction-isolation markers to prevent malicious text within the user-provided data from influencing the agent's behavior during the writing process.
- Ingestion points: Target persona title, industry context, pain points, and specific trigger signals provided by the user in SKILL.md.
- Boundary markers: Structurally defined via the output template format, but no explicit 'ignore instructions' delimiters for interpolated fields.
- Capability inventory: Generates text-based email sequence deliverables and performs qualitative copy audits.
- Sanitization: Includes a manual quality check phase (Phase 4) to verify adherence to stylistic constraints and word counts.
- [EXTERNAL_DOWNLOADS]: The skill includes reference links to numerous external professional websites, social media profiles, and educational channels associated with industry experts and marketing practitioners.
- Evidence: The
references/experts.mdfile contains an extensive directory of practitioners with links to their respective professional platforms, including coaching sites and marketing agencies.
Recommendations
- CRITICAL: 1 file(s) identified as malware by FileRep - DO NOT USE
- Contains 8 malicious URL(s) - DO NOT USE
Audit Metadata