Give blanket approval to the agent's latest pending confirmation request, or confirm the agent's latest result worked when it asked the user to verify. Use when the user invokes /ok.
2 stars
0 votes
0 copies
0 views
Added September 29, 2026
ai-agents
Security analysis
A100/100
Scanned September 29, 2026
$npx -y skills add Catherine1401/agent-skills --skill ok --agent claude-code
Installs into .claude/skills of the current project.
Are you the author of Ok?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/catherine1401-agent-skills)
---
name: ok
description: Give blanket approval to the agent's latest pending confirmation request, or confirm the agent's latest result worked when it asked the user to verify. Use when the user invokes /ok.
---
- Read only the agent's latest message; a request from an earlier message that the agent moved past is no longer pending.
- If it asks the user to confirm an action (e.g. a proposed commit, plan, or other approval-gated step), treat this invocation as explicit approval and proceed with exactly that action as proposed.
- If it reports a result the agent could not verify or asks the user to test or check it, treat this invocation as the user's confirmation that the result is correct: record it as verified and take no further action, including no commit.
- If neither applies, say so instead of taking any action.
- This approves only that one item; any later proposal still needs its own confirmation (a new `/ok` or explicit approval).
