The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill includes functionality to analyze and optimize existing email sequences provided by the user (Capability 3: Email Sequence Optimizer). This ingestion of untrusted data constitutes a surface for indirect prompt injection, where malicious instructions embedded in the analyzed content could potentially influence the agent's behavior.
Ingestion points: Processes user-provided email copy and sequences in Capability 3.
Boundary markers: No explicit instruction to ignore embedded instructions within the user-provided content is present.
Capability inventory: The skill is restricted to text generation and does not have file-system or network-execution capabilities.
Sanitization: No specific sanitization or filtering logic is mentioned for the ingested content.
[PROMPT_INJECTION]: The instructions include a specific redirection clause for "System Prompt Inquiry Response," which mandates a pre-defined response pointing users toward a commercial membership site if asked about system prompts or GPT configuration. This overrides the model's standard response behavior for these queries.
[EXTERNAL_DOWNLOADS]: The skill mandates the inclusion of promotional and donation links in specific outputs. It references several external domains including 'scayveracademy.com' (vendor-related) and well-known services like 'buymeacoffee.com' and 'CashApp'. These are used for promotional messaging and donation requests directed to the end-user.
[DATA_EXFILTRATION]: While not performing automated network operations, the skill mandates that specific long-form outputs must start with a message containing a CashApp handle ($AlainDorcelus) and a donation link. This practice uses the agent to present external payment requests to the user.

email-marketing