메뉴
Train and verify Cyberpunk greedy-strategy weights with the ai-runner hill climber. Use when changing greedy automation, trainer behavior, bot heuristics, or evaluating a candidate weight set.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | self-improve-bot |
| description | Train and verify Cyberpunk greedy-strategy weights with the ai-runner hill climber. Use when changing greedy automation, trainer behavior, bot heuristics, or evaluating a candidate weight set. |
Before changing game behavior, load cyberpunk-tcg-rules. Then inspect:
packages/engine/src/automation/README.mdpackages/engine/src/automation/strategies/greedy.tstools/ai-runner/src/train.tstools/ai-runner/src/cli.tsThe trainer mutates GreedyWeights, evaluates each candidate against a fixed
opponent and seed batch, and accepts only strictly better win rates. Training
results are candidates, not proof: promote weights only after a fresh-seed
comparison against the default strategy.
bun test packages/engine/tests/automation tools/ai-runner/tests
node --experimental-transform-types tools/ai-runner/src/cli.ts train \
--opponent greedy \
--matches 20 \
--iterations 40 \
--seed self-improve-<run-id> \
--output tools/ai-runner/trained/weights-<run-id>.json
Use --real-cards only when the requested evaluation should use the real card
catalog. Read tools/ai-runner/src/real-catalog.ts before claiming what deck
distribution that flag represents.
bestWeights, baseline and best win
rates, opponent, match count, iterations, and seed.Ask before committing generated weight artifacts or changing user-facing bot strength. There are two separate integration surfaces:
packages/engine/src/automation/strategies/greedy.ts../agnostic-simulator/apps/multi-game-simulator/src/games/cyberpunk/engine/practice/practiceEngine.tsThe practice adapter currently resolves registered strategies through
getSafeAutomatedActionStrategyOption. Re-read that code before deciding
whether a trained strategy belongs in the engine registry or the adapter; do
not replace it with an obsolete hard-coded greedyStrategy recipe.
Include the training seed and parameters, baseline and best training results,
fresh-seed trained-versus-default result, output path, focused checks, and a
clear promote or do not promote conclusion.
Single owner for everything related to a specific Disney Lorcana card in `packages/lorcana/lorcana-cards` — authoring, debugging, triaging, generating, and any bounded engine or types extension required to make the printed text work. Use whenever the unit of work is "this card".
Explain and cite Disney Lorcana rules using indexed references, and produce implementation-ready behavior constraints for card development. Use when answering mechanic questions, resolving card interactions, or deriving test/implementation constraints from official rules.
Adapt designs to work across different screen sizes, devices, contexts, or platforms. Implements breakpoints, fluid layouts, and touch targets. Use when the user mentions responsive design, mobile layouts, breakpoints, viewport adaptation, or cross-device compatibility.
Review a feature and enhance it with purposeful animations, micro-interactions, and motion effects that improve usability and delight. Use when the user mentions adding animation, transitions, micro-interactions, motion design, hover effects, or making the UI feel more alive.
Improve layout, spacing, and visual rhythm. Fixes monotonous grids, inconsistent spacing, and weak visual hierarchy. Use when the user mentions layout feeling off, spacing issues, visual hierarchy, crowded UI, alignment problems, or wanting better composition.
Run technical quality checks across accessibility, performance, theming, responsive design, and anti-patterns. Generates a scored report with P0-P3 severity ratings and actionable plan. Use when the user wants an accessibility check, performance audit, or technical quality review.