kami-balance-analysis-toolkit

Measures game balance decisions with persona-driven simulation and signed pacing bands.

1|Updated Jun 29, 2026
One-click install
npx skills add https://github.com/Raynos/kami-kakushi --skill kami-balance-analysis-toolkit-raynos
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: kami-balance-analysis-toolkit
Source: https://github.com/Raynos/kami-kakushi/tree/main/.claude/skills/kami-balance-analysis-toolkit
Command: npx skills add https://github.com/Raynos/kami-kakushi --skill kami-balance-analysis-toolkit-raynos

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Game balance changes in the kami-kakushi incremental RPG must be proven with simulation data rather than eyeballed, and this Skill encodes the full methodology: running the persona-bot sim, regenerating the committed pacing report, attributing unexplained report drift, and routing RED band violations to a human decision instead of fudging tests. ## Core Features & Use Cases - ADR-132 commit protocol: Runs verify:balance, regenerates docs/content/t0-pacing.md, and embeds a --summary block in the commit body so every magnitude change ships with its measured before/after. - Signed-band enforcement: Reads human-signed pacing constants from balance.ts, treats any RED band as a human decision fork, and explains the ADR-187 skip-blindness tradeoff of the greedy persona. - Worktree attribution proof: Regenerates the pacing report from a past SHA in a throwaway git worktree to prove which change moved unexplained t0-pacing.md rows. - Use Case: A designer asks whether a new cooking mechanic is too slow; you sim both worlds, find the rung wall-time exceeds the signed band by 6.6 minutes, and file an HD-item with the measured numbers instead of shipping or silently tuning. ## Quick Start Ask the assistant to run the balance simulation and check whether the current pacing report is fresh before proposing any change to a game balance value.

Frequently Asked Questions about kami-balance-analysis-toolkit

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I verify a game balance change with a simulation?

Run pnpm run verify:balance to execute the persona-bot sim against the real game engine, then regenerate the committed report with pnpm run balance:report and inspect the git diff of t0-pacing.md. Paste the output of pnpm run balance:sim --summary into the commit body.

What should I do when a pacing band check goes RED?

A RED band is a human decision, never a test-side fix. Either surface the fork as an HD-item with the measured numbers, or re-derive the design inputs such as requirement counts so the measurement moves; only the human may re-sign or widen a band.

How do I find out which commit changed the pacing report?

Create a throwaway git worktree at the suspect SHA, install dependencies, regenerate the report there, and diff it against the current report. Rows identical across both regenerations were moved by earlier commits; rows only your tree moves belong to your change.

Does the balance sim model all player behavior?

No. By standing ruling ADR-187 the greedy persona is deliberately skip-blind and never learns the sleep verb, so the sim under-represents sleep-heavy play. Real attended playtime from project/telemetry/ is the instrument that captures what the sim misses.

When should I not use this balance methodology?

Do not use it for fixing RED verify gates, learning game math formulas, adding new balance constants, or judging whether a feature is fun. The sim measures pacing, closure, and economy shape only; fun is assessed separately with the human.