Riddle

Hosted browser automation API for agents. Screenshots, Playwright scripts, workflows — no local Chrome needed.

Before installing: 1) Note the inconsistency: registry metadata lists no required credentials but the instructions require you to set a RIDDLE_API_KEY in OpenClaw config — confirm where the key is stored and how it is used. 2) Do not add the plugin to plugins.allow or edit ~/.openclaw/openclaw.json until you have inspected the plugin package (npm @riddledc/openclaw-riddledc) or its GitHub repo and verified checksums/CODE. 3) Understand privacy risks: sending cookies/localStorage/headers to a third-party browser service will expose session tokens and possibly PII — only do this with throwaway or explicitly consented credentials. 4) Verify the network allowlist and claims (api.riddledc.com only, no access to conversation history) by reviewing the plugin source and runtime policies; prose claims in SKILL.md are not proof. 5) If data sensitivity is high, prefer running Playwright locally or on an environment you control, or use ephemeral credentials and minimal scopes. 6) If you proceed, test with minimal, non-sensitive examples first and confirm the package's provenance (npm publisher, GitHub repo, CHECKSUMS.txt/SECURITY.md) and contact [email protected] with any questions.

Type: OpenClaw Skill Name: riddle Version: 1.1.0 The skill bundle is classified as benign. The `SKILL.md` provides clear instructions for installing and configuring a browser automation plugin, including setting an API key and modifying the OpenClaw configuration file (`~/.openclaw/openclaw.json`) using `jq` to add the plugin to an allow list. While direct file modification and external plugin installation (`@riddledc/openclaw-riddledc`) are present, they are explicitly explained as necessary for the plugin's functionality. Crucially, the `SKILL.md` includes a 'Trust & Security' section that declares strict boundaries for the plugin, such as network access limited to `api.riddledc.com`, filesystem writes restricted to `~/.openclaw/workspace/riddle/`, and zero access to agent context or other secrets. There is no evidence of prompt injection attempts, data exfiltration instructions, or other malicious intent within the provided files.