The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content from a local documentation file (docs/standards/tech-stack-rule.md) to determine the structure of generated Dockerfile and docker-compose.yml files. This creates a potential surface where malicious instructions in the project documentation could influence the resulting code.
Ingestion points: Reads technology stack definitions from docs/standards/tech-stack-rule.md and directory structures from docs/standards/directory-rule.md.
Boundary markers: There are no explicit instructions or delimiters defined to prevent the agent from obeying instructions that might be embedded within the project documentation.
Capability inventory: The skill is designed to write multiple configuration files (Dockerfile, docker-compose.yml, .dockerignore) to the project root.
Sanitization: No specific validation or filtering of the input documentation content is implemented.
[DATA_EXPOSURE_&_EXFILTRATION]: The skill demonstrates safe credential handling by explicitly instructing the agent to use .env.example templates and ensuring that sensitive information is passed through environment variables rather than being baked into Docker images.
[PRIVILEGE_ESCALATION]: The skill promotes the principle of least privilege by specifying that containers should run as a non-root user and that builder stages should be separated from runtime stages using multi-stage builds.

docker