Evidence-first game operations incident investigation agent with read-only MCP tools and reproducible evaluations.
# Add to your Claude Code skills
git clone https://github.com/nickname21kmr/gameops-investigatorGuides for using ai agents skills like gameops-investigator.
gameops-investigator is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by nickname21kmr. Evidence-first game operations incident investigation agent with read-only MCP tools and reproducible evaluations. It has 52 GitHub stars.
gameops-investigator's catalog security scan is still queued. You can run an instant dependency and prompt-injection check now with the "Scan for vulnerabilities" button above.
Clone the repository with "git clone https://github.com/nickname21kmr/gameops-investigator" and add it to your Claude Code skills directory (see the Installation section above).
gameops-investigator is primarily written in Python. It is open-source under nickname21kmr on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh gameops-investigator against similar tools.
No comments yet. Be the first to share your thoughts!
Unlocks once the catalog security scan passes (runs nightly).
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM is an informational directory and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
The deep catalog scan for this skill is still queued. Run an instant dependency check now instead.
English overview — An evidence-first game operations incident investigation agent built around metric contracts, read-only SQL, cohort comparison, anomaly testing, and citation-checked reports. The repository uses synthetic data; 17 automated tests and a 40-case deterministic offline evaluation validate the tool, safety, replay, and evidence pipeline—not Claude model quality.
Architecture · Browser demo · Security boundary · Onboarding · Evaluation cases
一个可运行、可审计、可评测的游戏运营异常归因 Agent。它把 Claude Code 作为规划与解释层,把指标、SQL、分群比较、异常检测和引用校验留给确定性程序。
数据声明:仓库内是固定种子生成的 5,000 名合成玩家和 136,164 条源事件。三个事故仅注入派生数据库,不代表真实商业游戏指标。


指标告警 -> Claude/回放协调器制定排查计划 -> MCP 工具调用 -> SQL 与分群下钻 -> 原因 Top-3 + 证据 -> 人工复核报告
内置三个可复现案例:
leveraged 玩家分群流失,整体检验未越过阈值但分群显著。system_opened 重复上报,事件级强度虚高而玩家级采用率基本稳定。工作台包含调查结论、完整工具轨迹、只读 SQL 沙盒、维度下钻图、40 条固定评测以及安全/泛化说明。
.\setup.ps1
.\run.ps1
浏览器打开 http://localhost:8501。setup.ps1 使用本机 F:\anaconda\python.exe(找不到时回退到 python)创建项目级 .venv,安装依赖、重建事故数据、生成三份报告、运行测试和 40 条评测。
如果环境已配置,只启动界面:
.\.venv\Scripts\python.exe -m streamlit run app.py --server.port 8501
仓库根目录的 .mcp.json 会注册 gameops stdio 服务器。setup.ps1 会在仓库 .tools/ 下安装便携式 Claude Code(不改系统 PATH);首次使用时完成一次交互式登录,然后:
.\claude-local.ps1 mcp list
.\.venv\Scripts\python.exe -m gameops_investigator.cli claude "Investigate the tutorial failure alert."
Claude Code 可调用五个工具:
get_metric_definition:指标口径、埋点、负责人和质量注意事项。query_metrics:受校验的只读 SQL,表白名单、行数和超时限制。compare_cohorts:版本、渠道、活动或玩家分层的确定性比较。detect_anomalies:两比例 z 检验或对数率比检验。draft_incident_report:带 SQL、证据 ID、置信度、限制和人工复核状态的报告。没有 Claude 登录也不影响演示、测试或评测:Streamlit 默认使用同工具链的 Deterministic replay。界面不会把这一路径冒充成 LLM。
.\.venv\Scripts\python.exe -m pytest
.\.venv\Scripts\python.exe evals\run_evals.py
默认结果写入 artifacts/eval_results.json。这里的分数是 deterministic offline baseline:验证固定题集、工具函数、回放协调器、安全策略和引用校验,不是 Claude 模型分数。
记录项包括:
Claude 的工具选择、归因、单次成本和延迟只有在完成已认证运行后才填写;Pro 订阅不被换算或冒充 API 成本。
核心代码不依赖 Newton 的玩法文案。接入另一款游戏需要:
config/schema_contract.json 的 users / events 最小模型;config/metrics.json 添加留存、漏斗、用户采用率、事件强度或数据质量指标;config/scenarios.json 添加告警窗口和下钻维度;查询防线、MCP 接口、统计检测、证据 ledger、报告引用校验和评测框架不需改写。详见 接入指南 与 安全边界。
app.py Streamlit 工作台
gameops_investigator/ 核心包与 MCP 服务
config/ 指标、场景和数据契约
data/source/ 不可改源快照
scripts/ 数据注入、报告和架构图生成
prompts/ 规划、指标解释、报告模板
evals/ 40 条固定评测与评分器
reports/ 三个事故报告和完整 trace
docs/ 架构、安全、接入与演示脚本
tests/ 单元与集成测试
可确认的表述:
基于 5,000 名合成玩家、136,164 条事件构建游戏运营异常归因 Agent,将指标查询、版本/用户分群对比、异常检测和报告生成封装为只读 MCP 工具,支持带证据的异常归因与 Text2SQL。
准确率、延迟和 Claude 成本请从本次实际 artifacts/eval_results.json 与已认证运行记录填写,不要手写数字。
这是一个 AI-assisted 个人作品:需求定义、指标口径、MCP 工具设计、异常案例、评测体系和结果复核均属于项目交付范围。请勿将确定性离线评测描述成 Claude 模型实测,也不要将合成数据描述成商业游戏数据。
data/source/ 中的合成数据采用 CC BY 4.0。.tools/、.venv/、Claude 登录状态及本地密钥不会进入版本库。