The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes docker build and docker run to verify generated Dockerfiles. This is the primary function of the skill and is constrained to local verification only.
[PROMPT_INJECTION]: The skill ingests codebase files which could contain indirect prompt injections. Evidence Chain: 1. Ingestion points: package.json, requirements.txt, go.mod. 2. Boundary markers: none. 3. Capability inventory: docker build, docker run. 4. Sanitization: alphanumeric/regex validation for image names, port numbers, and file paths.
[SAFE]: The skill follows security-first principles such as multi-stage builds, non-root user configuration, and prohibiting hardcoded credentials or latest tags.

dockerfile-generation