tevansandClaude Opus 5
07148df8e6
Build steps 1, 2 and 4 of the plan in DESIGN.md. core/ is pure JS with zero Foundry globals and 11 passing invariants. The fiction model has one root draw — an APPETITE plus the ACCOMMODATION someone made for it — and projects the motif, danger, prize, faction, cue fragments and the appeasement move from it, so the fiction coheres instead of being four unrelated draws. 336 kernels from 21 authored lines; 60 generated delves gave 51 distinct kernels. Cue fragments are tagged by facet (institution/ritual/demand/wound/anchor) and dealt without replacement. The facet progression across the arc reproduces the hand-authored paper delve's structure, which had arrived at it by instinct. Generation and table dice stay apart: pressure.mjs decides when a clock roll is due and what it means, and never rolls. That is what makes the skeleton seed an honest promise rather than a half-truth. paper/ holds the hand-written delve and both playtest logs. spike/ holds the instrumented harness that measured the Forge outside Foundry. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
8.9 KiB
PLAYTEST 02 — The Vault of the Gilded Court v2 (contrarian)
Pass: 2 of 2 (contrarian: Court Door · nobody bows · ally the gang · take the portrait) Scenario version: v2, absorbing playtest-01 fixes 1–8 Result: PASSES. Runnable. Four fixes to carry into v2.1, none blocking.
1. Method
Same instrument as pass 1 — same four pregens with real compendium numbers, real bestiary blocks,
CSPRNG for every contested beat, damage model per vanity.mjs:163. Contrarian choices taken at
every fork, including the ones v1 punished hardest.
Instrument gap, found mid-run: my opposed-roll helper did not check for Stumbles, so combat
Banes went uncounted. I re-scanned the log by hand and found two (Pinch [1,4,1,3,1] in A2,
Bram [2,4,1,1] in A6). Corrected below. Pass 1's Bane count was likely understated for the same
reason — its headline finding still stands, but "one Bane" should read "one Bane plus any combat
Stumbles I failed to log".
2. Play log
A1 — Court Door · ~10 min
Contrarian: the grand entrance. Clock rolled immediately per the attention trigger:
1d6 [1] → Something comes.
The attention rule works. The party announced itself and the barrow answered within a minute of play. That is exactly the intended cause-and-effect, and it fired on the very first roll.
Fix 4 held: the chisel-mark clue arrived automatically, so the motif seeded even though the party was in a hurry. In pass 1 this clue died on a roll.
A2 — Nobody bows · ~25 min
Three Skeletons (def 2 after fix 5). Wren carried it — 3d6[5,5,6]=3 net 2, then net 1,
killing two. Pinch stumbled: 5d6[1,4,1,3,1] — 0 successes, three 1s → +1 Bane.
⚠ Bram missed again: net −1, net −1. Across both passes he has now landed 2 hits in 12 attack rounds. Fix 5 lowered Ghoul and Skeleton defence and it was not enough — the problem is his 4d6 pool, not the opposition.
A3 — Ally the gang · ~20 min
Reaction 2d6 [2,1] → Hostile, again. The contrarian plan to ally was simply not on offer.
Fix 7 worked exactly as intended — the gang fought one round and broke on Nerve 2. Twenty minutes, not forty-five. This is the single clearest win of the pass.
⚠ But Hostile is the modal result. 2d6 needing 5–6 gives ~44% Hostile, and it has now
come up in both passes. The scenario offers "ally, rob, or abandon" as though allying is a live
option; nearly half the time the dice remove it before the players speak.
Yrsa lingered at a mirror (contrarian) → noticed → +1 Bane. The mirror rule is now a working Bane source, as fix 3 intended.
A4 — One Ghoul · ~15 min
Round 1: both missed. Round 2: Pinch rolled 5d6[5,5,5,4,5]=4 → net 4 → destroyed it outright.
Fixes 5 and 8 worked. Pass 1's two-Ghoul, defence-3 version ran 35 minutes and left one untouched. This ran 15 and resolved cleanly.
A5 — Take the portrait · ~20 min
The door trap fired (gang was hostile, so unwarned): Bram 4d6[2,5,2,5] → 2 successes, clear.
The Cantor's silvered blade was found — fix 1's insurance.
Portrait taken. The Queen woke here, on her ground.
⚠ Sim inconsistency: my run notes said "allied, so warned" while the reaction had been Hostile. The roll was made and passed either way, so the outcome is unaffected — flagging it for honesty.
A6 — The Queen, in the Wardrobe · ~40 min
Nobody had bowed and the tab was low, so she offered a Vice — fix 2's fallback. The contrarian party refused, and the fight started. That is exactly the branch v1 killed people with.
| Round | Queen Grit | Yrsa |
|---|---|---|
| 1 | 6 → 4 | untouched |
| 2 | 4 → 2 | −1 Grit, −1 max |
| 3 | 2 → 1 | −3 Grit, −1 max |
| 4 | 1 | −1 Grit, −1 max |
| 5 | 1 → 0 | — |
The Queen died in five rounds. Bram contributed two hits with the silvered blade. Bram also
stumbled in round 5 ([2,4,1,1]) → +1 Bane.
Fix 1 is confirmed. Pass 1: unwinnable, ~40 rounds, dead mage, Queen at 9/10. Pass 2: won in 5, with the mundane fighter participating.
⚠ Yrsa still goes down. She took 5 Grit against a maximum of 3, plus 3 permanent max Grit drained. Defence 2 against a 5d6 attack that ignores armour is not survivable, and every hit also shrinks her permanently. Lowering the Queen's Grit made the fight winnable but did nothing for the mage.
⚠ Spells deal 0 base damage, confirmed twice more: Yrsa scored net-1 hits in rounds 1 and 3 and dealt nothing both times. Her only real contribution was round 5's net 2. A mage in a straight fight is close to a bystander.
Total: ~2h10.
3. Player-hat verdicts
| Archetype | Verdict | Pass 1 | Pass 2 |
|---|---|---|---|
| Methodical veteran (Bram) | "The silvered blade finally let me into the boss fight. I still can't hit anything else." | 1/5 | 3/5 |
| Roleplay-first talker (Pinch) | "One roll deleted the Ghoul. Less to talk to this time — my own fault for not bowing." | 5/5 | 4/5 |
| Rules-literate tactician (Wren) | "I killed two skeletons and mattered in the finale. This is a real character now." | 2/5 | 4/5 |
| Lore-digger (Yrsa) | "Both clues arrived free, which was lovely. Then I did no damage for four rounds and got drained to nothing." | 1/5 | 2/5 |
Average 2.25 → 3.25. The scenario now works for three of four archetypes.
4. Fix list for v2.1
1. Yrsa dies in every finale. — high Problem: Defence 2 vs a 5d6 armour-ignoring attack that also drains max Grit. Winnable fight, dead mage. Fix: The Queen attacks whoever last looked at her — a stated, thematic targeting rule the table can play around, instead of defaulting to the squishiest. Add: "she turns to whoever last met her gaze." Where: Area 6 stat line.
2. Hostile is the modal reaction, so "ally" is rarely offered. — high Problem: ~44% Hostile, and it landed in both passes. The three-way choice is often a one-way. Fix: Give Pell Quill +1 die on the reaction if the party arrived by the Servants' Stair (they didn't announce themselves) and let a hero spend 1 Vanity to re-roll it as a parley. Where: Area 3 reaction block.
3. Bram cannot hit anything. — high Problem: 2 hits in 12 attack rounds across both passes. Fix 5 was insufficient. Fix: This is a pregen problem, not a scenario problem — Bram's 4d6 is simply low for a fighter. Give the scenario's own §2b a note: the warrior should open with Cleave or a Called Shot rather than trading attacks, and put a shield-wall option in A2 so he has a job that isn't rolling to hit. Where: §2b; Area 2.
4. The tab reached 3, not 6. — medium Problem: Banes: Pinch stumble, Yrsa mirror, Bram stumble = 3. The Reckoning still did not fire — though it is now plausibly reachable, where in pass 1 it was not. Fix: Add +1 Bane for entering by the Court Door (announcing yourself is vanity) and +1 for refusing the Queen's Vice. Both are pure character moments and would have put this party on 5. Where: Areas 1 and 6; §9.
5. What this says about DELVE (draft 4/5)
| # | Finding | Consequence |
|---|---|---|
| 1 | 6 areas ran ~2h10 once the fixes landed | Pass 1's "6 is too many" was wrong — it was 2h55 because the content was broken, not because six is too many. Withdraw the recommendation to drop to 4–5. Codex's caution was reasonable but the evidence now says 6 is fine at this heat. |
| 2 | The clock fired on the first roll, from a player choice | Attention-triggered clocks work far better than turn-count clocks. DELVE should weight triggers over the 3-Turn timer. |
| 3 | acceptanceCost: null is not a temptation — it is an automatic yes |
In both passes every free-acceptance item was taken without discussion. The decision only ever appeared at use. Draft 5 should stop describing acceptance as the temptation and put the whole economy at the point of use — or give a minority of relics real acceptance friction. This answers scenario question 3, finally. |
| 4 | A "way back in" rescued the branch that killed pass 1 | The Vice offer meant a party that ignored the intended route still had an ending. DELVE's generated areas should always emit a fallback route, not just a gate. |
| 5 | Two content traps came from stat blocks, again | Confirms the overlay catalog needs a playability field. Add a second: "is this fight survivable by the party's weakest member?" — both passes lost the mage. |
6. Status
The Vault of the Gilded Court v2 is runnable. Apply the four v2.1 fixes before a live table; none of them block a run today.
Build-order step 1 is complete: one handwritten delve, run twice, findings folded back. The scenario went from unwinnable to won-in-five, and produced five structural findings for the design that no amount of further architecture work would have surfaced.
Next: step 2, the seam spike.
Private VANITY development material — not for sale or distribution.