The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data fetched from external OCI APIs, which presents a surface for indirect prompt injection if resource metadata (such as instance names or image descriptions) is controlled by a malicious actor.
Ingestion points: The skill retrieves external data using compute.list_instances, compute.list_shapes, and compute.list_images in SKILL.md.
Boundary markers: The instructions do not define clear boundaries or use specific delimiters to isolate potentially untrusted API data from the agent's core instructions.
Capability inventory: The skill has access to shell execution (Bash) and powerful OCI management tools that can create, modify, or delete cloud infrastructure.
Sanitization: There is no evidence of sanitization or schema validation for the data returned by the OCI service before it is presented to the agent.
[DATA_EXFILTRATION]: The skill accesses the local file ~/.oci/config, which contains sensitive authentication credentials including OCIDs and key fingerprints. While this is a standard operational requirement for the official Oracle Cloud SDK to authenticate requests, the access to credential files is a sensitive operation.

oraclecloud-hello-world