Compare strategy versions under fixed game conditions
Using Claude where available: Supply strategy versions, opponents, random seeds and scoring rules to design a consistent evaluation.
Execution environment and usage
AI / software
General text chat (example: Claude); original stack retained below
Where to operate
A general text conversation, for example a new chat at claude.ai where available. This is the suggested exercise entry, not an attribution of the original tool.
Getting started
- Open claude.ai or another text-chat tool you are authorized to use.
- Convert the required material below into readable, source-labeled text and paste only necessary authorized excerpts.
- Enter the case prompt and check the resulting draft against the original material.
Setup and plan requirements
Synthetic strategy versions, opponents, seeds and scoring. Actual rerun needs harness, container image/model setup and compute budget. Text chat requires an available account and allowance; Claude has location and age requirements. Manually transcribe images or scans first rather than assuming text chat can inspect the original file.
Prompt portability limits
Plan does not run matches; historical standings do not generalize to current models.
Usage limitations
This review performed no installation, hands-on test or regional-availability check. Agent Setup section describes isolated Docker containers, OpenCode and tournament orchestration.
Environment sources
- https://llmskirmish.com/screeps · 2026-10-19
- https://support.claude.com/en/articles/8114491-get-started-with-claude · 2026-10-19
適用情境 / When to use
Use the plan to check whether improvements hold beyond selected matches. Retain script failures and missing results. Running matches needs the game harness, agent tools and isolated environment; the plan itself runs nothing. Results apply to that study’s models, harness and tournament design.
步驟 / Steps
- Prepare synthetic strategy versions, opponents, seeds and scoring rules.
- Hold evaluation conditions fixed and compare outcomes and invalid scripts.
- Retain failures and sample limits rather than selecting favorable matches.
提示詞 / Prompt
Using [synthetic strategy versions/opponent field/random seeds/scoring rules/test budget], design a revision comparison. Fix conditions and separate valid matches, script failures and missing results, reporting sampling limits and possible overfitting. Plan simulation evaluation only; do not enter live matches, bet or claim universal model superiority.
留意事項 / Cautions
- Do not connect real competitive accounts or place bets.
- Do not impose the source’s round count or model ranking.
來源限制 / Evidence limits
Standings are not current model rankings; more history does not guarantee a better strategy.
流程說明 / Workflow note
These steps and prompts are original, untested exercises by this site, not the author’s verbatim workflow. Reported results do not guarantee exercise outcomes.
原始工具組合 / Original tool stack
OpenCode; isolated containers; multiple language models; Screeps-style simulation
Source publication date: 2026-02-04
Source checked: 2026-10-20