# Flex Call

> A game-morning fantasy football lineup board. Paste a roster with your own projections, floors, ceilings, injury tags (Q, D, O, IR) and notes: the browser applies a fixed tag discount, fills every lineup slot (standard, 3 WR, 2 FLEX or superflex), benches byes and OUT players, honours players you pin or keep out, and lists the close calls and the flags - free. Then ask for an injury triage (play, monitor or pivot for every tagged player) or for the close calls settled (who starts in each coin flip, with high, medium or low confidence). Every reply cites the board's players, slots, close calls and flags by id and is checked against them.

URL: https://flex-call.skillsafe.ai/
API: https://flex-call.skillsafe.ai/api.html
Built on the SkillSafe skills @onewave-ai/fantasy-lineup-optimizer (sit/start decisions with confidence levels) and @onewave-ai/injury-report-tracker (fantasy impact of injuries, backup options, return timing).

## What runs free in the browser (flexkit.js, board.js)

- Roster parsing: one player per line - name, position, team, opponent (@ABC, vs ABC or BYE), projection, then optionally floor and ceiling, a tag and a note. Commas, tabs, pipes or a pasted Markdown table; .csv/.txt files by picker or drag-and-drop. Players get ids P1, P2, ...
- Board points: your projection x 1 (healthy), x 0.85 (Q), x 0.35 (D), x 0 (O, IR, bye). This is the page's rule of thumb, not a forecast.
- Slot filling: strict slots first, then FLEX (RB/WR/TE), then SUPERFLEX (QB/RB/WR/TE) - optimal for nested eligibility. Pins (tap a bench player, then a slot) and kept-out players are honoured.
- Close calls (D1, D2, ...): for each unpinned slot, the best bench challenger within max(2 points, 15%) of the starter, one per challenger and per starter, at most six.
- Flags: empty slot, questionable or doubtful starter, byes, OUT/IR players, missing projections, your defence facing your own players, QB-receiver stacks, a pinned player who cannot score, no healthy backup for a hurt starter.
- Exports: Markdown, lineup CSV, JSON and a printable lineup card.

## The two metered tasks (model gpt-terra)

- `triage` - verdict `all_clear`, `watch_list` or `scramble`; for each tagged player `play`, `monitor` or `pivot`, a replacement from your roster, when to check again, and why; bench depth notes. One click benches the pivots and hands the board to the close calls.
- `calls` - verdict `keep`, `tweak` (one swap) or `rework` (two or more); for each close call the player to start (only the two names in it), a confidence level and why; a stack/conflict note; risks. One click pins the calls on the board.
- Both: no live news, practice reports, weather or betting lines - only what the roster says; numbers only from the board; every line cites ids. `FlexKit.reconcile` checks that every tagged player or close call is answered exactly once, that O/IR/bye players are pivoted, that every pivot is on the roster, playable and slot-compatible, that every start is one of the two names, that the verdict matches the calls, and that every cited id and every decimal number is on the board.

## Notes

- Sample players, teams and projections are invented. Not affiliated with any league or fantasy platform. Not betting advice.
- Saved lineups go to the signed-in user's own collection (`lineups`), only when they press save.