The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill facilitates the execution of system commands based on external data, creating a vulnerability surface. 1. Ingestion points: User-provided image names, build contexts, and command arguments for docker run and docker exec (SKILL.md). 2. Boundary markers: Absent. No recommendation for delimiters or instructions to ignore embedded commands is provided. 3. Capability inventory: High-risk capabilities including docker build, docker run, and docker exec are used throughout the skill. 4. Sanitization: No sanitization of user-provided strings is documented.
[Command Execution] (SAFE): The provided shell commands are instructional examples for Docker management and do not perform unauthorized or hidden actions.
[Credentials Unsafe] (SAFE): The docker-compose.yml file uses generic placeholders like secret for passwords, which does not constitute a credential leak.

docker-management