| name | history-course-qa |
| description | Verify scare-the-spire History Course seeded replay behavior against real local Slay the Spire 2 `.run` history files. Use when Codex changes `src/lib/sts2-run-replay.ts`, `/history-course` replay/map/topbar behavior, run upload parsing, build-version RNG branching such as v0.107.1 MegaRandom, or when QA needs evidence that uploaded runs still reconstruct exact act paths. |
History Course QA
Overview
Use this skill to validate History Course replay reconstruction with the user's actual local STS2 run history. The canonical smoke is scripts/survey-all-runs.ts, which reads Steam .run files and reports exact, ambiguous, and zero-match act reconstructions by build.
Source Data
The current survey script reads:
~/Library/Application Support/SlayTheSpire2/steam/76561199168753671/profile1/saves/history
If that directory does not exist, inspect scripts/survey-all-runs.ts and the local ~/Library/Application Support/SlayTheSpire2/steam/*/profile*/saves/history tree before changing the script. Do not commit user .run files.
Workflow
- Inspect the changed scope:
git status --shortgit diff --name-only HEAD- recent commits from the current task if the working tree is clean.
- Run the survey:
pnpm exec tsx scripts/survey-all-runs.ts
- Compare results against the relevant baseline from the task, not just the total:
- Before the v0.107.1 RNG fix, local survey was
160 runs / 303 acts / EXACT 223 / ZERO 80.
- After the v0.107.1 MegaRandom branch, expected local survey is about
EXACT 302 / ZERO 1, with v0.108.0 zero 0.
- Treat any new ZERO in pre-v0.107.1 builds as a likely regression unless explained by a separate map rule change.
- Treat any new ZERO in v0.107.1+ builds as a likely RNG, map-generation, or run parsing regression unless the run spans a game patch boundary.
- For a failing row, open the corresponding
.run file from the local Steam history directory and inspect:
build_id, seed, acts, ascension, players, modifiers- act variants: standard, Spoils Map, Golden Compass, Winged Boots, Big Game Hunter, multiplayer
- whether the file was likely started before and finished after a patch.
- If a replay change is meant to fix a failure, rerun the survey after the fix and report the per-build delta.
Required Checks
Run these for History Course replay logic changes:
pnpm exec tsc --noEmit --pretty false
pnpm lint
pnpm exec tsx scripts/survey-all-runs.ts
Also run targeted Playwright specs when UI behavior or screenshots can regress:
pnpm exec playwright test scripts/history-course-shell.spec.ts
pnpm exec playwright test scripts/history-course-rewards.spec.ts
Use $mobile-viewport-qa when route layout, topbar density, map rendering, tooltip surfaces, or responsive History Course UI changed.
Reporting
Report the exact command output summary, especially total runs, acts, exact, ambiguous, zero, and per-build zero counts. Include any known residual failure with filename, build, seed, and the likely reason.
