The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the run_command tool to execute docker exec and docker compose commands, which grants the agent significant control over the local Docker environment.\n- [COMMAND_EXECUTION]: The troubleshooting section contains destructive commands such as docker compose down --volumes --rmi all which remove persistent data and images without adequate warning.\n- [INDIRECT_PROMPT_INJECTION]: The skill uses string interpolation for <asset_key> and <partition_date> in shell commands. If these inputs are sourced from untrusted data, they could be used to perform command injection.\n
Ingestion points: <asset_key> and <partition_date> variables used in SKILL.md.\n
Boundary markers: None present.\n
Capability inventory: Shell command execution via docker exec and docker compose.\n
Sanitization: No input validation or escaping is defined for the interpolated arguments.

dagster