Nextbrowser1

The name/description say the skill will spin up cloud browsers, manage profiles, and run autonomous browser tasks — the SKILL.md contains API calls for profiles, locations, credentials, and tasks that directly implement that purpose. However, the metadata declares no primary credential or required config paths even though the runtime instructions require an API key stored in openclaw config (skills.entries.next-browser.apiKey). That mismatch between declared metadata and runtime requirements is an inconsistency.

The SKILL.md instructs the agent to read the OpenClaw config for an API key, call Nextbrowser endpoints to list/store credentials and profiles (which persist logins/cookies), and to run autonomous 'task' subagents that can log into third-party accounts and perform actions (examples include upvoting and posting comments). It also explicitly recommends always using skip_plan_approval=true and 'fast' mode, which reduces manual oversight. These instructions give the agent broad authority to act on user accounts and to store/ reuse sensitive credentials; that scope goes beyond simple read-only browser inspection and could enable abusive behavior if misused.

This is an instruction-only skill with no install spec and no code files, so it does not write binaries or download third-party code during install. That minimizes install-time risk.

Although the skill needs an API key, the registry metadata lists no required environment variables and no config paths. The SKILL.md requires an API key stored at openclaw config path skills.entries.next-browser.apiKey and shows examples using an $API_KEY header. The omission of any declared primary credential or required config path in the metadata is a mismatch. Additionally, the skill instructs use of Nextbrowser-managed credential IDs (i.e., storing/using third-party account credentials), which is high-sensitivity functionality and should be clearly declared and justified.

The skill is not marked always:true and does not request elevated install persistence. However, it recommends skip_plan_approval=true for autonomous subagents, which effectively reduces approval checkpoints and increases the risk of automated, unchecked actions. The skill also relies on Nextbrowser to persist profiles and credentials outside the agent — that persistence is a functional requirement but a privacy/abuse risk and should be considered when enabling the skill.