The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied application requirements and external documentation to generate Dockerfiles, creating a standard surface for instructions that could influence generated output.
Ingestion points: Application requirements collected in Stage 1 and external framework documentation fetched via documentation lookups or web search in Stage 2.
Boundary markers: The skill instructs the agent to extract only actionable data from research, though it does not define specific prompt delimiters for user input interpolation.
Capability inventory: The skill allows writing files to the local file system (Dockerfile, .dockerignore) and executing shell-based validation tools.
Sanitization: A mandatory validation-iteration loop (Stage 6) is implemented, requiring the use of security scanners such as hadolint and checkov to audit and repair the generated artifacts.
[COMMAND_EXECUTION]: The skill utilizes shell scripts and external binaries for the generation and verification of container configurations.
Generation Scripts: Reference implementations in the scripts/ directory use shell utilities and sed to interpolate user requirements into templates.
Validation Tools: The workflow integrates static analysis tools including hadolint, checkov, and a custom validator script to audit the security posture of the generated Dockerfiles.
[EXTERNAL_DOWNLOADS]: The skill performs downloads from well-known and trusted sources during its execution.
Documentation Retrieval: The skill uses context7 and web search tools to retrieve production deployment best practices for specific frameworks.
Dependency Resolution: The generated templates include standard instructions for fetching dependencies from official registries such as NPM, PyPI, and Linux distribution repositories (Alpine/Debian).

dockerfile-generator