The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the datahub CLI tool to perform metadata searches and lineage traversals. The instructions in SKILL.md (Step 1) explicitly mandate input validation by instructing the agent to 'Reject shell metacharacters in search queries and URNs before passing to CLI,' which effectively mitigates common command injection risks when interacting with shell-based tools.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external sources (DataHub metadata) and user-supplied strings which are then incorporated into subsequent CLI commands.
Ingestion points: User-supplied entity names and URNs; metadata and lineage graphs returned by the datahub search and datahub lineage commands as described in SKILL.md.
Boundary markers: The instructions provide clear operational boundaries for the agent (e.g., 'Not This Skill' section) and structural templates for output, though explicit delimiters for external data strings are not required for this use case.
Capability inventory: The skill is restricted via allowed-tools to the Bash(datahub *) environment, limiting its execution capability to the specific vendor toolset.
Sanitization: The skill includes explicit sanitization instructions to check for and reject shell metacharacters in user input prior to interpolation into bash commands.

datahub-lineage