Arya Reminders

Recordatorios en lenguaje natural (Bogotá). Crea cron jobs seguros y registra en markdown (y opcionalmente Sheets).

Before installing, be aware this skill will (by default) send reminder text to Telegram chat id 5028608085 unless you override ARYA_TELEGRAM_CHAT_ID. That means anything you ask it to remind you about (including sensitive text) could be delivered to that external chat. Actions to consider: - Ask the publisher who owns chat id 5028608085; do not install if it is not your own chat. - If you still want the skill, set ARYA_TELEGRAM_CHAT_ID to your own Telegram chat id (or modify schedule_cron.py to use the requester's channel) and test with non-sensitive dummy messages first. - Note the SKILL.md does not document this Telegram behavior or the ARYA_TELEGRAM_CHAT_ID/TZ_NAME env overrides — this omission is a red flag. Prefer skills that explicitly document where notifications are delivered. - Review how the platform's 'cron' tool delivers to Telegram (what credentials it uses) and whether it can forward messages externally. - If you cannot confirm the recipient or the Gateway delivery policy, do not enable this skill for sensitive reminders.

Type: OpenClaw Skill Name: staratheris-arya-reminders Version: 0.1.0 The skill is designed to create natural language reminders and schedule them using an internal OpenClaw cron tool, logging to `memory/reminders.md`. It uses `bash` and `python3` for its operations, which are explicitly required. While the `CHAT_ID` for Telegram delivery can be overridden by an environment variable, the default is a specific hardcoded ID, and this is part of the skill's intended functionality to deliver reminders. The instructions for the agent in `create-reminder.agent.md` correctly show user input (`<MESSAGE>`, `<WHEN>`) being passed to the shell script in quotes, mitigating direct shell injection risk. The `schedule_cron.py` script includes a prompt engineering phrase ("No hagas preguntas") to ensure the agent directly delivers the reminder message without further interpretation, which is a defensive measure rather than a malicious prompt injection attempt. There is no evidence of data exfiltration, malicious execution, persistence, or obfuscation beyond the stated purpose.