The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a bash script template that uses curl to interact with the ClickHouse HTTP interface and tar to archive diagnostic results. These commands are intended for local diagnostic collection.
[EXTERNAL_DOWNLOADS]: The skill references the @clickhouse/client Node.js package in its documentation. This is a well-known, official library for ClickHouse connectivity.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface (Category 8) by reading from system.query_log. This table contains raw SQL queries executed by database users, which could contain adversarial instructions intended to influence the agent during the analysis of the diagnostic bundle.
Ingestion points: Query logs are retrieved from system.query_log in the manual SQL steps, the bash script (Step 5), and the Node.js collector (Step 6).
Boundary markers: None present; data is collected as raw TSV or JSON output.
Capability inventory: The agent has access to Bash, Read, and Grep tools which could be targeted by instructions found in logs.
Sanitization: No filtering or escaping of the retrieved query text is performed prior to presenting it to the agent environment.

clickhouse-debug-bundle