The Agent Skills Directory

[External Dependencies]: The skill installs the official google-api-python-client and google-auth Python packages. These are standard, well-maintained libraries provided by the vendor for interacting with Google Cloud APIs.
[Command Execution]: The skill uses gcloud and curl to interact with Google Cloud services. It follows the principle of least privilege by focusing on read-only diagnostics and mandates explicit user confirmation before executing any state-changing commands or modifying host configuration.
[Indirect Prompt Injection]: The skill processes external data from Cloud Logging and API responses to perform diagnostics. While this is necessary for its intended use, it creates a surface where external content could influence agent reasoning. The skill includes specific safety and confirmation rules to mitigate potential risks associated with autonomous actions.
Ingestion points: External data is retrieved via gcloud logging read and curl requests to GCP REST APIs (documented in SKILL.md and references/logging-and-monitoring.md).
Boundary markers: The instructions do not specify explicit delimiters for external data, but provide a structured flow for the agent to follow.
Capability inventory: The agent can execute shell commands, perform network requests, and manage local Python packages.
Sanitization: The skill does not explicitly instruct the agent to sanitize or escape data retrieved from logs before processing.

application-integration-troubleshooting