Skill
Discernment Nudge
A small instruction skill that prompts a verification pause without connecting to any account or data source.
Published by the company that makes the product.
What it can access
No account, file, or network permission of its own. The host assistant still has whatever permissions and data access you give that conversation.
Security note
Use it as a pause, not a fact-check. Read the SKILL.md before installation and remove it if its extra questions distract from low-risk work.
Do you need it?
Use it if participants act too quickly on fluent answers. Skip it if the workflow already includes a clear verification checklist.
Capability and permissions
The skill adds instructions that ask what to check before acting. It has no account, file, or network permission of its own. It is not a fact-checker.
Least-privilege setup
As of 2026-09-01, obtain the skill only from Anthropic’s official repository. Read SKILL.md, package only that skill folder, and upload it through the current Skills settings in Claude. Do not add tools or account credentials.
Harmless success test
Ask for advice about an invented public event. Confirm the answer ends with useful verification questions and does not claim those questions have been checked.
Disable, revoke, remove
Disable or delete the skill from the assistant’s Skills settings. It creates no external authorisation to revoke.
When not to connect
Do not treat it as evidence, source checking, legal review, or protection assessment. Remove it when the extra prompt obscures a simple low-risk task.