OpenClaw Paid Actions

Use the openclaw_paid_action tool to list actions, generate USDC invoices, and execute only after manual payment confirmation on Solana.

What to check before installing: 1) Confirm you have a trusted implementation of the openclaw_paid_action tool (source, release, or vendor) because the SKILL is instruction-only. 2) Review every configured action command (e.g., scripts/paid-actions/*) before enabling; those commands execute with the agent's privileges and receive the action input via OPENCLAW_PAID_ACTION_INPUT_JSON. 3) Store OPENCLAW_PAID_ACTIONS_INVOICE_SECRET securely and consider using a signing key with limited scope/funds for testing. 4) Ensure invoice store path is on a secure filesystem and that the agent's config storage is trusted. 5) Limit tools.allow and do not enable autonomous invocation unless you trust the configured actions and have enforced reviewed-scripts policy. 6) If you need higher assurance, ask the publisher for source code or a release URL and verify the openclaw_paid_action implementation before enabling the skill.

Type: OpenClaw Skill Name: openclaw-paid-actions Version: 0.2.2 The skill bundle is classified as suspicious due to the described architecture in `SKILL.md` that allows for the execution of external scripts (e.g., `scripts/paid-actions/x-shoutout.mjs`) with input derived from potentially untrusted sources (`OPENCLAW_PAID_ACTION_INPUT_JSON`). While the `SKILL.md` itself does not contain malicious code or direct prompt injection, it outlines a design pattern where a vulnerability in the external scripts could lead to Remote Code Execution (RCE). The documentation acknowledges this risk by recommending `enforceReviewedScripts: true` and advising to 'Review every configured action command before enabling autonomous execution', indicating a known attack surface.