The Agent Skills Directory

[DATA_EXFILTRATION]: The scripts/prose.ts script reads text from standard input and the docs/korean-style.md file, then sends this content to the OpenRouter API endpoint (https://openrouter.ai/api/v1/chat/completions). This is the primary data flow for the prose correction service.
[CREDENTIALS_UNSAFE]: The script programmatically loads the OPENROUTER_API_KEY from a .env.local file in the current working directory. This implementation automates the retrieval of credentials for the API service.
[PROMPT_INJECTION]: The skill processes external data that is interpolated into the model's prompt, creating a surface for indirect prompt injection.
Ingestion points: Content is read from process.stdin and docs/korean-style.md in the current working directory.
Boundary markers: The prompt uses --- and 참고 자료: as delimiters between instructions and user-provided data.
Capability inventory: The skill uses fetch for network communication and readFileSync for file access.
Sanitization: No specific sanitization or validation logic is applied to the input text to prevent malicious instructions from influencing the model's behavior.
[SAFE]: No obfuscation, unauthorized command execution, or persistence mechanisms were detected. The script uses standard Node.js built-in modules.

korean-prose