The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from agent traces and logs, creating a potential surface for indirect prompt injection if malicious content is present in the logs being analyzed. Ingestion points: Data enters the agent context through CLI commands like adk logs and adk traces as described in references/traces-and-logs.md and references/conversation-analysis.md. Boundary markers: The instructions lack explicit delimitation or instructions to ignore embedded commands within the analyzed trace data. Capability inventory: The skill performs shell command execution using the platform's adk CLI and accesses project configuration files. Sanitization: No explicit sanitization or escaping mechanisms are provided for handling the ingested data before summarization.
[COMMAND_EXECUTION]: The skill makes extensive use of platform-specific CLI commands (adk check, adk logs, adk traces, etc.) to retrieve diagnostic information. These are standard, legitimate operations for a debugger skill intended for developers.

adk-debugger