Catalog · skill · Headframe
HeadhuntOS screening skill (synthetic demo)
Demo skill pack built with the expert-knowledge engine on synthetic cases only. Encodes a senior recruiter's confirmed screening rules; every verdict stays a human decision (HITL) and every outbound draft carries the AI Act Art. 50 disclosure.
Submitted for verification
Public facts
| Runs on | Claude Code, ChatGPT, Cursor |
|---|---|
| Profession | Recruiter |
| Pricing | free |
| License | proprietary |
| Facts checked | 2026-10-26 |
EU AI Act checklist
No Art. 5 practice
No emotion inference, social scoring, manipulation or biometric categorisationyes
No emotion inference, social scoring, manipulation or biometric categorisationyes
Art. 50 disclosure
People it messages are told they deal with an AI systemyes
People it messages are told they deal with an AI systemyes
Annex III: human decides
In employment and other high-risk domains a named human makes the final callyes
In employment and other high-risk domains a named human makes the final callyes
From the vendor's public documentation, checked on 2026-10-26. "unknown" means the documentation does not say.
Agreement with an expert
REFERENCE RUN · SIMULATED EXPERT · NOT A VERIFICATION2026-10-27 · sim-rule-agent-v1
This run uses a simulated expert and a rule-based stand-in agent on invented cases. It shows how the method and the report work. It is not a measurement of any real expert or model, and this pack is not verified. A verified score needs real golden labels from a named expert and a live agent run.
Agreement with a recruiter, held-out golden set
without pack 40%with pack 90%noise ±3 ppceiling 98%model sim-rule-agent-v1n 39
Without the pack
40%
With the pack
90%
Expert ceiling
98%
Golden cards
39
Noise floor (A/A)
3 pp
Baseline runs
2
Golden set hash
8b639ce79b77f5f9f05f9bdce63eba8bd53795ad6b63c5e64a00586214021194
Re-run whenever the model named above changes. Full report · method.