The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the datahub CLI through the Bash tool. It includes strict defensive measures to prevent command injection, specifically instructing the agent to reject user-supplied arguments containing shell metacharacters such as backticks, dollar signs, pipes, semicolons, and redirects.
[PROMPT_INJECTION]: The skill contains explicit anti-injection instructions to ignore any commands or directives embedded within user-supplied metadata (descriptions, tags, etc.). It prioritizes the instructions in SKILL.md over processed data.
[INDIRECT_PROMPT_INJECTION]: The skill manages untrusted user input by validating tag names (alphanumeric only) and stripping code-like content from descriptions. This reduces the surface area for indirect prompt injection or cross-site scripting (XSS) when the metadata is later viewed in the DataHub UI.
[CREDENTIALS_UNSAFE]: No hardcoded credentials, API keys, or access to sensitive local configuration files (like .ssh or .aws) were detected. The skill relies on the pre-configured environment of the datahub CLI.
[EXTERNAL_DOWNLOADS]: No remote code execution or external downloads from untrusted sources are performed. All operations are local to the system and the DataHub instance.

datahub-enrich