The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill generates Oracle APEX artifacts based on functional requirements and data models provided as external input (Markdown or JSON). This ingestion of untrusted data constitutes an indirect prompt injection surface. The skill mitigates this risk by requiring a "Live Runtime Validation Gate" using the official Oracle apex validate tool. All generated code must pass this authoritative compiler check before the agent is permitted to import it into a database, effectively sanitizing the output.
[COMMAND_EXECUTION]: The skill makes extensive use of localized tooling (node tools/apexctl.mjs) and the Oracle SQLcl command-line interface (sql, apex) to perform project discovery, artifact validation, and deployment. These command executions are fundamental to the skill's stated purpose and are restricted to the local environment and configured Oracle database targets.
[DATA_EXFILTRATION]: The skill performs network requests using Node.js http and https modules to verify the status and reachability of the generated APEX application's runtime environment. These requests are limited to the user-provided or environment-defined APEX_RUNTIME_BASE_URL and are used exclusively for legitimate diagnostic and verification purposes.
[CREDENTIALS_UNSAFE]: The skill handles database connection aliases (db_connection_name) to interact with Oracle databases. It includes explicit instructions to never log secrets or full connection strings, ensuring that sensitive credentials are not exposed in logs or session transcripts. The provided templates use placeholder values for salts and identifiers rather than hardcoded secrets.

apex