The Agent Skills Directory

[DATA_EXFILTRATION]: The instructions in SKILL.md include code that accesses the sensitive OCI configuration file at ~/.oci/config. This file typically contains API signing keys, tenancy OCIDs, and user OCIDs used for authentication.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from OCI Monitoring and Logging services, which could be a vector for indirect prompt injection attacks.
Ingestion points: Data is ingested via MonitoringClient.summarize_metrics_data, MonitoringClient.list_metrics, and LogSearchClient.search_logs methods in SKILL.md.
Boundary markers: The provided scripts do not utilize delimiters or specific instructions to isolate the retrieved monitoring data from the agent's internal logic.
Capability inventory: The skill uses the oci Python SDK for network communication and service management, and it is permitted to use Bash(pip:*) for environment setup.
Sanitization: There is no evidence of sanitization or validation of the retrieved log entries or metric data before it is processed or displayed by the agent.

oraclecloud-observability