PocketLens

Use when user wants to track expenses, scan receipts, upload card payment screenshots, categorize spending, record transactions, check spending summaries, vi...

This skill appears to do what it says: it uses Node and your PocketLens API key to upload parsed receipt/statement data to pocketlens.app. Before installing: (1) only provide an API key you trust and give it the minimum necessary permissions (the README requests a write/full key to create transactions), (2) do not set POCKET_LENS_API_URL to an unknown host — that would send your API key elsewhere, (3) be aware uploaded images will be processed by the agent's vision tool (they may be sent to the platform's vision provider), and (4) inspect the included script if you want extra assurance (it is readable and uses only fetch to call /api/external/* endpoints). Revoke the key if you stop using the skill or if you suspect misuse.

Type: OpenClaw Skill Name: pocket-lens Version: 1.0.0 The skill integrates with PocketLens for expense tracking, which is a benign purpose. However, a significant vulnerability exists in the `SKILL.md` instructions for the OpenClaw agent. Specifically, when the agent is instructed to execute `node pocket-lens.mjs create-transaction '<JSON>'`, the JSON argument is derived from user input (either from image analysis or manual entry). If the OpenClaw agent fails to properly sanitize or shell-escape this user-controlled JSON string before passing it to the shell, it could lead to shell injection, allowing arbitrary command execution on the host system. While the skill's instructions suggest using single quotes for the JSON argument, the ultimate responsibility for robust escaping of user-controlled content lies with the agent's implementation, making this a high-risk vulnerability rather than intentional malice within the skill itself.