The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a POSIX-compliant shell script located at scripts/verify-docker-stack.sh. This script is intended to be executed via the run_command tool to audit project directories for compliance with the Zero-Host-Dependencies standard. The script performs local file system discovery using find and inspects file contents using grep to validate Docker and Makefile configurations.
[INDIRECT_PROMPT_INJECTION]: The skill exposes an attack surface for indirect prompt injection via its verification script.
Ingestion points: The scripts/verify-docker-stack.sh script reads local project files, specifically Makefile, Dockerfile*, docker-compose.yml, and docker-compose.dev.yml (located in the user-specified target directory).
Boundary markers: None identified; the script performs raw pattern matching using grep without explicit delimitation for the agent context.
Capability inventory: The script uses find, grep, and printf. The agent context utilizes run_command to execute the audit and view_file to read the reference guides.
Sanitization: None; the script processes file contents directly to detect specific strings (e.g., USER, FROM, help:).
[SAFE]: The skill demonstrates high security awareness by providing detailed guidelines for container hardening, including the enforcement of USER 1000:1000 to avoid root execution, the use of --ignore-scripts to mitigate supply chain attacks during dependency installation, and the recommendation of distroless or scratch base images for production environments.

docker-and-orchestration