The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources such as CRM and payment platforms, creating an attack surface where malicious content in those records could attempt to influence the agent's behavior. \n- Ingestion points: External data is retrieved via the ~~crm and ~~платежи references. \n- Boundary markers: The instructions do not specify any delimiters or safety prompts to isolate external data from the agent's core instructions. \n- Capability inventory: The skill utilizes the Bash, WebFetch, and Read tools to process the retrieved data. \n- Sanitization: There is no evidence of data validation or sanitization being performed on the input from external sources.

friday-brief