Read AI

Fetch and manage Read AI meeting data - summaries, transcripts, action items, and engagement metrics. Use when the user asks about meetings, meeting notes, m...

This skill appears to do what it says: it queries Read AI / Limitless APIs and saves meeting data under your home directory. Before installing, consider: 1) the API key is stored in a plain file (~/.config/readai/api-key) — protect that file (SKILL.md suggests chmod 600). 2) The Limitless pull optionally calls a local 'claude' CLI when you use the --ai flag — only enable that if you trust the 'claude' binary on your system. 3) The webhook receiver defaults to localhost but can be bound to all interfaces; only bind publicly if you know what you’re doing and secure the endpoint (use a firewall, reverse proxy with TLS, or tunnel). If those caveats are acceptable, the skill is internally consistent with its purpose.

Type: OpenClaw Skill Name: readai Version: 1.0.0 The skill bundle contains several security vulnerabilities and risky patterns that could be exploited via prompt injection. Most notably, `scripts/limitless_pull.py` is vulnerable to path traversal because it uses the command-line `date` argument to construct filesystem paths without sanitization, potentially allowing arbitrary file writes. Additionally, `scripts/readai_client.py` provides an export feature that allows writing data to arbitrary file paths via the `--output` flag, and `scripts/webhook_receiver.py` implements an unauthenticated network listener that binds to a local port. While these appear to be functional features or unintentional flaws rather than evidence of malicious intent, they represent a significant attack surface.