Website ScreenshotOne (by ScreenshotOne)

This skill appears to do what it says (build curl calls to ScreenshotOne), but there are a few practical risks and inconsistencies to weigh before installing: - The SKILL.md requires SCREENSHOTONE_ACCESS_KEY but the registry metadata does not declare any required env vars — confirm the skill owner and update metadata or refuse installation until the required credential is declared. - The examples pass the access_key as a query parameter. Query parameters (and generated URLs) can be logged or appear in Referer headers; prefer using a secure header-based flow or at least be aware of leakage risk when using sensitive URLs or shared environments. - The documented options include metadata_content and storage_* settings. If you enable metadata_content=true, the API can return page content (potentially sensitive). If you set storage_endpoint/storage_bucket, screenshots could be uploaded to arbitrary storage — ensure you control the destination. - Because this is an instruction-only skill with allow_implicit_invocation enabled, an agent could call it automatically when relevant. If you are concerned about automatic captures of sensitive URLs, disable implicit invocation or only use the skill in sessions where you explicitly permit it. What to do before installing: verify the skill owner (the registry owner id), insist the registry metadata be corrected to list SCREENSHOTONE_ACCESS_KEY as a required credential, and avoid using the skill with pages that contain sensitive data unless you control the ScreenshotOne account and any storage destinations. If you need more assurance, ask the publisher to change examples to use header-based auth or POST body auth to reduce URL leakage.