The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests external data in the form of PingOne audit logs via the searchAuditActivities and getAuditActivity tools (referenced in references/playbooks/audit-execution.md). To mitigate risks, it implements boundary markers by rendering data in markdown tables and raw JSON blocks as defined in references/catalogs/audit-output.md. The capability inventory is restricted to read-only information retrieval tools like listUsers and listApplications (found in SKILL.md), ensuring that the agent cannot execute unauthorized state-changing commands. Sanitization is not explicitly implemented, but the risk is managed by the clear demarcation of external data from agent instructions.

pingone-platform-ops