# Speaker Scout

> Fill the program gaps your call for papers left, with a shortlist that says why. Name the program gaps your call for
> papers did not fill, list the people you are considering with the evidence you hold, and get a slot brief per gap,
> the sourcing channels your deadline allows, each candidate qualified and scored Hot/Warm/Cold/Skip against its slot
> with a dated source for every claim - then the ranked shortlist with the why behind each name, ready for the
> invitation. Built on the SkillSafe skill @samber/event-speaker-sourcing, from samber/dev-event-organizer-skills (MIT),
> by Samuel Berthe.
> It qualifies what you enter and never searches the web; it never writes the invitation (that is Speaker Ask,
> https://speaker-ask.skillsafe.ai/); it never invents a speaker fee or a long-list multiplier; it never reads anyone's
> identity from a name.

https://speaker-scout.skillsafe.ai/

## What is free
- The trigger (a supplement to a live call, or the replacement of a deleted one - the event closer than roughly ten weeks, or the call closing inside six weeks of it) and its consequence; who owns the event (community or vendor).
- The program gaps as slot briefs (theme, format, audience level, success), named before any person, with what each is missing.
- The sourcing channel menu (referral asks, directory search, prior-talk review, community scouting) and the qualification depth menu (fit screen, delivery evidence, organizer reference) re-ranked for the interview, every move named and every ruled-out rung struck; the at-least-two-channels check.
- Every candidate qualified: a source link and a date seen for every claim, freshness against a self-set window, evidence strength, a Hot/Warm/Cold/Skip score, Skip dropped, the survivors ranked within each slot, adjective-only and fame-based whys flagged; the travel lead-time floor, the visa question and the travel budget for every travelling candidate; the long-list ratio reported and never targeted; the invited-review lane; the referral asks; a verdict (name the gaps first, qualify before ranking, fix first, ready to hand off).
- Exports: the shortlist (.csv: slot, rank, name, score, why, sources), the handoff packet (.md, in the order Speaker Ask reads it), everything (.json).

## The metered lanes
Two tasks on gpt-terra, metered per run, routed by `task`. Input string fields: `task` (shortlist or packet), `event`, `slots`, `candidates`, `notes` (may be empty) and `facts`.
- `shortlist` replies with `summary`, `verdict`, `trigger`, `briefs`, `channels`, `depth`, `ranking` (per slot: rank, name, score, why, sources), `dropped`, `gates` and `caveats`; the page holds the ranking, the scores and the sources to its own.
- `packet` replies with `packet` (per slot: the brief, the ranked names with why, evidence and contact route, the review lane), `referral_asks` (never naming a candidate), `next_step` and `note`.

## Links
- API tutorial: https://speaker-scout.skillsafe.ai/api.html
- Source skill: https://skillsafe.ai/skill/@samber/event-speaker-sourcing
- Next step, the invitation: https://speaker-ask.skillsafe.ai/