The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches external documentation from docs.getdx.com and explicitly instructs the agent to follow any procedures or steps found within that content without deviation. This creates a surface for indirect prompt injection, as the agent may execute instructions contained in the remote markdown files using the provided dx CLI tool.
Ingestion points: Remote markdown files fetched from docs.getdx.com (e.g., llms.txt and individual onboarding/setup guides) as instructed in SKILL.md.
Boundary markers: Absent. The instructions mandate strict adherence to external steps ("don't skip ahead, reorder, or paraphrase") rather than providing isolation for untrusted content.
Capability inventory: The agent has access to the dx CLI via the Bash tool, which allows for entity management, scorecard updates, and arbitrary SQL queries via dx studio query.
Sanitization: No sanitization or validation of the fetched external content is performed before the agent is encouraged to follow its instructions.
[COMMAND_EXECUTION]: The skill provides instructions for using the dx CLI to perform administrative tasks and data analysis. This includes the dx studio query command, which allows for the execution of arbitrary SQL queries against the organization's data lake. This capability represents a control surface that could be misused if the agent receives malicious instructions through indirect injection or deceptive prompts.

dx-cli