The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands to validate the generated Dockerfile, specifically calling docker build, docker run, docker inspect, docker logs, hadolint, and trivy. These commands are used to verify the application's containerization and log output.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from the project structure and source code to derive build commands and dependencies.
Ingestion points: The agent reads project manifest files (e.g., package.json, go.mod, requirements.txt), source files, and environment configuration files (e.g., .env.example).
Boundary markers: No explicit delimiters or instructions are provided to the agent to treat external codebase content as untrusted or to ignore embedded instructions within those files.
Capability inventory: The skill allows the agent to execute docker build, docker run, and docker logs using parameters derived from the ingestion points.
Sanitization: There is no evidence of sanitization or filtering of the content extracted from the codebase before it is used in shell commands or Dockerfile generation.
[REMOTE_CODE_EXECUTION]: The skill performs local code execution via docker build and docker run. Since the logic and commands within the Dockerfile are dynamically generated from the user's codebase, an attacker could potentially execute malicious commands within the container during the validation phase.

generate-dockerfile