The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user data via command-line arguments for resource group IDs and labels.
Ingestion points: Arguments in scripts/create_resource_group.py and scripts/patch_resource_group.py (e.g., --id, --label).
Boundary markers: None present; values are interpolated directly into API calls.
Capability inventory: The scripts execute network operations via the AICoreV2Client to create or modify resources in SAP AI Core.
Sanitization: Basic regex validation is performed on resource group IDs in scripts/create_resource_group.py (^[a-z0-9][a-z0-9\-]*[a-z0-9]$), but labels are not validated beyond the key=value format.
[DATA_EXFILTRATION]: The scripts/get_token.py script outputs a raw authentication Bearer token to the standard output.
Evidence: print(token) in scripts/get_token.py. If invoked directly by the agent without redirection to a variable, this sensitive token will be included in the agent's context and conversation history.
[CREDENTIALS_UNSAFE]: The skill is designed to manage and access sensitive SAP AI Core credentials.
Evidence: The scripts access configuration files in ~/.aicore/config.json and environment variables like AICORE_CLIENT_SECRET. It also supports X.509 certificate and private key authentication methods.
[PERSISTENCE]: The setup documentation instructs users to modify shell profile files to persist environment variables.
Evidence: references/SETUP.md suggests adding export commands to ~/.zshrc or ~/.bashrc to maintain credential configuration across sessions.

aicore-admin-resources