The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to execute a shell command (ls) on the host system to check for the presence of recommended companion skills in specific hidden directories (e.g., ~/.claude/skills/). This is used for its 'Companion check' functionality.
[EXTERNAL_DOWNLOADS]: The skill recommends the installation of additional data engineering tools via npx from the vendor's repository (AbsolutelySkilled/AbsolutelySkilled). These downloads originate from the skill's own author and are intended for project setup.
[PROMPT_INJECTION]: As an advisor for data infrastructure, the skill processes user-supplied SQL and Python code. While this represents a surface for indirect prompt injection, the skill does not contain instructions that bypass safety filters or exfiltrate data. Evidence includes ingestion of dbt models and Spark scripts from the user environment.

data-pipelines