The Agent Skills Directory

[SAFE]: The skill incorporates robust defensive instructions to protect sensitive data. It explicitly forbids the agent from requesting, printing, or forwarding authentication tokens and mandates the use of secure OAuth 2.1 flows. It also prevents privilege escalation by instructing the agent to never attempt to bypass role-based or tenant-based access restrictions.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data via MCP tools to summarize documents and audit trails, which introduces a potential attack surface for indirect prompt injection if those external records contain malicious instructions.
Ingestion points: Data is retrieved from the campus environment via tools including spmi_list_documents, spmi_get_document_detail, and spmi_get_audit_trail as defined in SKILL.md.
Boundary markers: The skill does not explicitly define delimiters or instructions to ignore embedded commands within the processed document content.
Capability inventory: The agent's capabilities are restricted to read-only monitoring; it lacks permissions for file modification, shell execution, or arbitrary network communication.
Sanitization: No specific content sanitization or validation procedures are listed for the data retrieved from external tools.

spmi-gpm