Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when 处理玩家可见的游戏 UI、玩法互动、引导、失败恢复、挑战链路或体验验收。
.claude/skills/kiakun-collab-player-interaction-design/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 73% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 46% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 74% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 110% | 0% |
本 skill 只触发玩家互动设计流程，不重复 AGENTS.md、任务包或项目 guide。
优先级：
AGENTS.mddocs/agent_guides/**如果冲突，按更高优先级执行，并说明冲突点。
优先读取当前项目内的同名规则；如果项目未提供对应文件，读取本 skill 打包资源：
references/player_interaction_contract.mdassets/templates/player_interaction_contract.mdscripts/validate_player_interaction_contract.py涉及玩法、UI、GDD、文案、引导、失败恢复或结算复盘，先读： docs/agent_guides/player_experience_first.md；若项目没有该文件，则按本 skill 的玩家互动合同执行。
涉及玩家互动合同，读： docs/agent_guides/player_interaction_contract.md；若项目没有该文件，读本 skill 的 references/player_interaction_contract.md。
涉及证据、截图、Playtest Bridge、visible-window 或体验结论，读： docs/agent_guides/evidence_levels_and_claims.md；若项目没有该文件，至少按本 skill 的证据层级选择表收口。
写完整合同：
写短合同：
如有疑问，默认写完整合同。宁可多写，不漏失败恢复路径。
实现、拆任务包或写验收前，先写玩家互动合同。
证据层级必须使用项目既有口径，并按 player_interaction_contract.md 的前置条件选择。只跑 headless/focused runner 时，不得声称玩家真实体验通过或自然通关。
证据层级 是结构化字段，只填写 logic_runner、ui_contract、interactive_mcp、visible_capture、manual_canary 或明确被任务要求的 natural UI-only。runner、fixture、加速、bridge helper、fallback 等支撑说明写到 证据说明、验收要求或 not_proven，不要塞进 证据层级。
写完合同或任务包后，运行：
shpython3 scripts/qa/validate_player_interaction_contract.py --mode auto <markdown...>
如果项目没有该脚本，改用本 skill 打包脚本：
shpython3 <skill-dir>/scripts/validate_player_interaction_contract.py --mode auto <markdown...>
常见误区见 docs/agent_guides/player_experience_first.md。
证据层级 写“目标最低为 ui_contract，可选补 visible_capture”。应写 ui_contract / visible_capture。natural UI-only；否定时写“自然 UI 多日链”或“自然通关”。| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 30,529 | 6,084 | -80% | 1 | 1 | 0% | 5,618 | 1,206 | -79% | 0 | 0 | — |
case-02 | fail→pass | 15,919 | 20,448 | +28% | 1 | 1 | 0% | 3,597 | 3,503 | -3% | 0 | 0 | — |
case-03 | fail→fail | 19,232 | 15,852 | -18% | 1 | 1 | 0% | 3,304 | 3,414 | +3% | 0 | 0 | — |
case-04 | fail→pass | 10,364 | 11,006 | +6% | 1 | 1 | 0% | 1,646 | 2,847 | +73% | 0 | 0 | — |
case-05 | pass→pass | 9,828 | 3,476 | -65% | 1 | 1 | 0% | 1,717 | 1,431 | -17% | 0 | 0 | — |
case-06 | fail→pass | 9,570 | 7,557 | -21% | 1 | 1 | 0% | 1,525 | 2,229 | +46% | 0 | 0 | — |
case-07 | fail→pass | 6,408 | 4,624 | -28% | 1 | 1 | 0% | 978 | 1,697 | +74% | 0 | 0 | — |
case-08 | fail→pass | 22,680 | 8,761 | -61% | 1 | 1 | 0% | 1,097 | 2,305 | +110% | 0 | 0 | — |
case-09 | fail→pass | 9,426 | 4,085 | -57% | 1 | 1 | 0% | 1,585 | 1,556 | -2% | 0 | 0 | — |
case-10 | pass→pass | 9,020 | 7,550 | -16% | 1 | 1 | 0% | 1,444 | 2,112 | +46% | 0 | 0 | — |
case-11 | fail→pass | 5,345 | 2,024 | -62% | 1 | 1 | 0% | 988 | 1,195 | +21% | 0 | 0 | — |
case-12 | fail→pass | 8,455 | 3,091 | -63% | 1 | 1 | 0% | 1,466 | 1,399 | -5% | 0 | 0 | — |
case-13 | fail→pass | 14,679 | 2,315 | -84% | 1 | 1 | 0% | 2,229 | 1,265 | -43% | 0 | 0 | — |
case-14 | fail→pass | 6,963 | 2,718 | -61% | 1 | 1 | 0% | 1,144 | 1,370 | +20% | 0 | 0 | — |
case-15 | fail→pass | 6,399 | 2,330 | -64% | 1 | 1 | 0% | 1,013 | 1,284 | +27% | 0 | 0 | — |
case-16 | fail→pass | 18,436 | 5,403 | -71% | 1 | 1 | 0% | 1,509 | 1,882 | +25% | 0 | 0 | — |
case-17 | fail→pass | 12,011 | 5,158 | -57% | 1 | 1 | 0% | 2,120 | 1,748 | -18% | 0 | 0 | — |
case-18 | fail→pass | 14,124 | 5,120 | -64% | 1 | 1 | 0% | 2,446 | 1,918 | -22% | 0 | 0 | — |
case-19 | fail→pass | 14,378 | 7,098 | -51% | 1 | 1 | 0% | 2,326 | 2,121 | -9% | 0 | 0 | — |
case-20 | pass→pass | 10,425 | 4,314 | -59% | 1 | 1 | 0% | 1,773 | 1,589 | -10% | 0 | 0 | — |
case-21 | pass→pass | 16,002 | 8,975 | -44% | 1 | 1 | 0% | 3,458 | 2,525 | -27% | 0 | 0 | — |
case-22 | pass→pass | 10,440 | 12,782 | +22% | 1 | 1 | 0% | 2,254 | 3,607 | +60% | 0 | 0 | — |
case-23 | pass→pass | 11,249 | 8,849 | -21% | 1 | 1 | 0% | 2,316 | 2,637 | +14% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted, and 21 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +65 percentage points is the difference between those two pass rates over the 21 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.