The Agent Skills Directory

[PROMPT_INJECTION]: The SKILL.md file contains a "Directive: EXECUTE, DON'T ASK" and "Enforcement Behavior" section. These instructions explicitly command the AI to bypass standard user confirmation prompts when executing commands, which is an override of core safety protocols.
[COMMAND_EXECUTION]: The skill is designed to execute shell commands via docker exec and docker compose. When combined with the directive to skip user permission, this enables the autonomous execution of commands without human oversight.
[EXTERNAL_DOWNLOADS]: Both README.md and SKILL.md provide instructions to download and install additional skills from external GitHub repositories, specifically wrsmith108/docker-enforce, wrsmith108/docker-optimizer, and wrsmith108/docker-guard.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface through its command transformation logic:
Ingestion points: User-supplied command strings (e.g., npm install <package>) described in SKILL.md and setup.md.
Boundary markers: Absent; there are no delimiters or instructions to ignore embedded commands within the user-provided package names or arguments.
Capability inventory: Execution of system commands via docker exec, docker compose, and bash across SKILL.md, setup.md, and health-checks.md.
Sanitization: Absent; the skill does not define methods for validating or escaping the user-provided input before incorporating it into execution strings.

docker