Scrask

When the user sends a screenshot via Telegram, parse it using Gemini (fast, default) with automatic Claude fallback when confidence is low. Saves results to...

This skill appears to do what it says: it sends screenshots to Gemini (and optionally Anthropic/Claude) for parsing and then writes events/reminders/tasks into Google Calendar/Tasks using a service account. Before installing: (1) Be comfortable with images and extracted text being sent to third‑party LLM providers (Gemini/Anthropic). (2) Create a least‑privileged Google service account, grant only the Calendar/Tasks scopes needed, and review which calendar is shared with the service account. (3) Verify the GEMINI_API_KEY and optional ANTHROPIC_API_KEY you provide are scoped/managed according to your policy. (4) Inspect the included scrask_bot.py locally (you have the source) to confirm there are no hidden network calls or logging of secrets beyond what you expect (the provided code appears to call only the model SDKs and Google APIs). (5) Note minor packaging/documentation inconsistencies (file paths and version number) — consider testing in a sandbox before enabling on a primary account.

Type: OpenClaw Skill Name: scrask-bot Version: 1.0.1 The skill is classified as suspicious due to a potential shell injection vulnerability in `SKILL.md`. The instructions for the AI agent include executing a `python3` command with interpolated configuration variables (`$CONFIG_VISION_PROVIDER`, `$CONFIG_TIMEZONE`). If the OpenClaw agent does not sanitize these user-configurable values before executing the shell command, a malicious configuration could lead to Remote Code Execution (RCE). The `scrask_bot.py` script itself appears benign, handling API keys securely via environment variables and performing its stated function of parsing images and interacting with Google APIs.