The Agent Skills Directory

[SAFE]: The skill uses official Ballerina ecosystem resources and tools (bal CLI, Ballerina Central, and official Docker images) which are standard for the development workflow described.
[CREDENTIALS_UNSAFE]: The coding rules explicitly instruct against hardcoding secrets, recommending the use of configurable variables and environment variables instead, which is a secure practice.
[DATA_EXFILTRATION]: The skill includes guidelines to redact sensitive data (secrets, auth tokens, raw payloads) from logs before outputting them.
[PRIVILEGE_ESCALATION]: The provided Dockerfile uses the --chown flag and a non-root user (ballerina), following the principle of least privilege for containerized environments.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data in the form of OpenAPI specifications (oas.yaml) to generate code stubs.
Ingestion points: The bal openapi command reads the oas.yaml file to scaffold services.
Boundary markers: Absent; there are no specific instructions for the agent to ignore malicious prompts within the YAML file.
Capability inventory: The skill utilizes bal openapi (code generation), bal build (compilation), and bal run (execution).
Sanitization: Absent; the skill does not specify validation or sanitization of the OpenAPI file contents before processing.

ballerina