The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/deploy.sh script employs the eval command within its run_command function to execute shell operations. The command strings are constructed using variables such as COMPOSE_FILE, STACK_NAME, and NAMESPACE. If these inputs contain shell escape characters, it could lead to arbitrary command execution.
[DYNAMIC_EXECUTION]: The use of eval in scripts/deploy.sh to process dynamically generated strings at runtime is a risk, particularly because the script targets files found within the user's project directory.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involve scanning the local project for docker-compose.yml and Dockerfile files to generate configuration. This creates a surface where malicious content within those project files could influence the agent's behavior.
Ingestion points: Project scanning in SKILL.md and file reading in scripts/generate_env_file.py.
Boundary markers: No delimiters or instructions are used to distinguish project data from agent instructions.
Capability inventory: Access to Bash (docker and kubectl) and Write tools.
Sanitization: There is no evidence of sanitization or content validation for the files read from the project.
[EXTERNAL_DOWNLOADS]: The scripts/deploy.sh script references a download URL for the kompose tool from the official Kubernetes GitHub repository. This is a reference to a well-known and trusted organization.
[CREDENTIALS_UNSAFE]: Multiple files including assets/compose_template.yml and assets/example_app_architectures.md contain hardcoded example credentials like POSTGRES_PASSWORD=password. While intended as placeholders for templates, these represent insecure defaults if used without modification.

generating-docker-compose-files