bfowle si-daily-challenge — Daily Spirit Island Drill
Generates a single daily challenge card for Brett. The selection logic lives in
generate.py (same directory) so the randomization, weighting, and playlog
reads are deterministic and testable instead of hand-rolled in prose.
How to invoke
Default daily run (date-seeded, stable within a single day):
python3 ~/.claude/skills/si-daily-challenge/generate.py
Pipe stdout into your response verbatim — it is already formatted as a markdown challenge card.
Re-rolls and overrides
Pass flags to vary the draw. The script already handles edge cases; prefer a flag over re-describing intent in prose.
| Brett says | Flag |
|---|---|
| "pick a different spirit" | --different-spirit |
| "pick a different adversary" | --different-adversary |
| "different goal" | --different-goal |
| "reroll" / "another one" | --reroll N (bump N each time) |
| "I want harder" | --harder |
| "easier" | --easier |
| "give me X spirit" | --spirit <slug> |
| "vs England" / "against " | --adversary <slug> |
| "at level N" | --level N |
| "solo" / "2-handed" / "3-handed" | `--players 1 |
| "no scenario" | --no-scenario |
| "with " | --scenario <slug> |
Multiple flags compose. If Brett re-runs /si-daily-challenge within the same
day, bump --reroll (1, 2, 3…) so the draw actually changes — the default
seed is the date.
Selection logic (reference — actual behaviour lives in generate.py)
- Spirit — weighted random across all 37 spirits. Priority flag adds +2, full chapter adds +2 (both stack). Recent-5 from playlog are excluded.
- Adversary — from Brett's go-to pool (England / Sweden / Brandenburg- Prussia / France-Plantation). Least-recently-played wins, random tiebreak.
- Level — walks that adversary's
difficulty_levelsladder. Default = lowest rung; +1 after 2 consecutive wins at the current rung against that adversary; −1 after 2 consecutive losses. - Scenario — 20% chance; weighted toward diff_mod ≤ 1.
- Board — one random board per spirit. 70% base (A–D), 30% expansion (JE E–F, HoSI G–H).
- Player count — 55% true solo, 30% 2-handed, 10% 3-handed, 5% 2-player.
- Learning goal — date-rotated through 7 categories;
--different-goalskips one slot.
Output shape
The script prints:
## Today's Challenge — YYYY-MM-DD
**Spirit**: …
**Adversary**: … Level …
**Scenario**: …
**Board**: …
**Player count**: …
**Learning goal**: …
### Read first (15 min)
- Spirit chapter: …
- Fundamentals: …
- Adversary chapter: …
### Turn-1 commit
Before you start, write on paper: "By T3 I will have …"
### After the game
Run `si-post-game` to log; tomorrow's challenge shifts based on results.
A stub-chapter footnote appears when the chosen spirit has no full chapter, pointing Brett at the spirit index + Wiki.
JSON mode
For programmatic callers: generate.py --json emits the raw selection as
JSON (spirit, adversary, level, scenario, boards, goal, player_count, seed).
Anti-alpha note
Never prescribe specific plays. The card names chapters and goals; the chapter has the turn-by-turn detail. Advisory, not prescriptive.
After output
Ask in one line: "Ready for this, or want me to swap the spirit / adversary / goal / level?" — then stop. Don't narrate further.