The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill documents and uses patterns where sensitive database credentials (username and password) are embedded directly in the URL for curl commands, such as https://USER:PASSWORD@HOST:443/. This is an insecure practice as credentials in URLs can be exposed in shell history, process listings, and network logs.
[COMMAND_EXECUTION]: The skill extensively uses the Bash tool to execute curl for database operations and the op (1Password) CLI to retrieve secrets from an engineering vault.
[EXTERNAL_DOWNLOADS]: The skill fetches sensitive data from 1Password's cloud service using the op command-line interface.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by interpolating variables retrieved from external tools (like op) directly into shell commands and SQL statements.
Ingestion points: Credentials and configuration data are ingested via the op item get command in SKILL.md.
Boundary markers: No specific delimiters or warnings are used to prevent the agent from interpreting instructions potentially embedded in the retrieved data.
Capability inventory: The skill uses Bash to execute curl and op commands, which can be manipulated if the input data is compromised.
Sanitization: There is no evidence of sanitization or validation of the hostnames or credentials before they are used in shell commands.

clickhouse-cloud-management