The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection (Category 8).
Ingestion points: The skill ingests untrusted data from ServiceNow tables including interaction_entry (chat transcripts), sys_journal_field (journal comments), and kb_knowledge (knowledge articles).
Boundary markers: There are no delimited structures or instructions provided to the agent to treat these external inputs as untrusted or to ignore embedded instructions.
Capability inventory: The skill utilizes Bash (native) and multiple ServiceNow-specific MCP/REST tools for data retrieval and potentially broader system interaction.
Sanitization: No logic is present to sanitize, validate, or filter the external content before it is interpolated into the generation prompt.
[DATA_EXFILTRATION]: The skill accesses sensitive Customer Service Management (CSM) data, including personally identifiable information (PII) such as customer names, email addresses, phone numbers, and account tiers from the csm_consumer and customer_account tables. While this access is aligned with the skill's stated purpose of providing personalized agent assistance, it involves the handling of sensitive record types.

chat-recommendation