Back to Details

cxas-agent-foundry

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Agent Foundry

Agent工厂

End-to-end lifecycle for GECX conversational agents: build, test, debug, iterate.
GECX对话式Agent全生命周期管理：构建、测试、调试、迭代。

Quick Reference

快速参考

bash
undefined
bash
undefined

Lint before push (catches structural issues early)

推送前进行代码检查（提前发现结构问题）

cxas lint --app-dir <project>/cxas_app/<AppName>
cxas lint --app-dir <project>/cxas_app/<AppName>

Push local files to platform

将本地文件推送到平台

cxas push --app-dir <project>/cxas_app/<AppName>
--to projects/<project_id>/locations/<location>/apps/<app_id>
--project-id <project_id> --location <location>
cxas push --app-dir <project>/cxas_app/<AppName>
--to projects/<project_id>/locations/<location>/apps/<app_id>
--project-id <project_id> --location <location>

Pull platform state to local

将平台状态拉取到本地

cxas pull projects/<project_id>/locations/<location>/apps/<app_id>
--project-id <project_id> --location <location> --target-dir <project>/cxas_app/
cxas pull projects/<project_id>/locations/<location>/apps/<app_id>
--project-id <project_id> --location <location> --target-dir <project>/cxas_app/

Run evals + triage + report (single command)

运行评估 + 故障分类 + 生成报告（单命令）

python .agents/skills/cxas-agent-foundry/scripts/run-and-report.py --message "what changed" --runs 5
python .agents/skills/cxas-agent-foundry/scripts/run-and-report.py --message "what changed" --runs 5

Inspect app architecture

检查应用架构

python .agents/skills/cxas-agent-foundry/scripts/inspect-app.py
python .agents/skills/cxas-agent-foundry/scripts/inspect-app.py

Triage failures

故障分类

python .agents/skills/cxas-agent-foundry/scripts/triage-results.py --last 3
undefined
python .agents/skills/cxas-agent-foundry/scripts/triage-results.py --last 3
undefined

Environment Readiness Check (run BEFORE routing)

环境就绪检查（在路由前运行）

Before routing to any sub-skill, check these signals in order:
  1. Virtualenv exists? -- Check if 
    .venv/
    directory exists
  2. Config exists? -- Check if 
    .active-project
    file exists and the referenced 
    <project>/gecx-config.json
    exists
  3. Has built before? -- Check if any 
    <project>/cxas_app/
    directory has content
SignalAction
No 
.venv/
or no config		First-time setup needed. Load 
references/setup.md
before doing anything else.
gecx-config.json
exists but no 
cxas_app/
content		Returning user, new project. Route normally.
All existReturning user. Route normally.
在路由到任何子技能之前，按顺序检查以下信号：
  1. 是否存在Virtualenv？ —— 检查是否存在
    .venv/
    目录
  2. 是否存在配置？ —— 检查是否存在
    .active-project
    文件，且引用的
    <project>/gecx-config.json
    存在
  3. 是否已构建过？ —— 检查是否有
    <project>/cxas_app/
    目录包含内容
信号操作
无
.venv/
或无配置		需要首次设置。在进行任何操作前先加载
references/setup.md
。
gecx-config.json
存在但
cxas_app/
无内容		返回用户，新项目。正常路由。
全部存在返回用户。正常路由。

Detect Intent and Route

意图检测与路由

Read what the user wants and load the appropriate sub-skill:
User says...PhaseLoad
"Build me an agent from this PRD"Build
references/build.md
"Create evals for my agent"Build
references/build.md
"Generate tool tests", "create callback tests"Build
references/build.md
"Update evals -- requirements changed"Build
references/build.md
"Update the TDD"Build
references/build.md
"Run evals", "push evals", "check results"Run
references/run.md
"Run tool tests", "test the callbacks"Run
references/run.md
"Generate a report"Run
references/run.md
"Why is this eval failing", "get to 90%"Debug
references/debug.md
"Fix the failing evals", "debug the agent"Debug
references/debug.md
"Tool test is failing", "callback test broke"Debug
references/debug.md
If the intent is unclear, ask: "Are you looking to build/create evals, run them, or debug failures?"
理解用户需求并加载相应的子技能：
用户表述...阶段加载内容
"根据这份PRD为我构建一个Agent"构建
references/build.md
"为我的Agent创建评估"构建
references/build.md
"生成工具测试"、"创建回调测试"构建
references/build.md
"更新评估——需求已变更"构建
references/build.md
"更新TDD"构建
references/build.md
"运行评估"、"推送评估"、"查看结果"运行
references/run.md
"运行工具测试"、"测试回调"运行
references/run.md
"生成报告"运行
references/run.md
"为什么这个评估失败了"、"达到90%的通过率"调试
references/debug.md
"修复失败的评估"、"调试Agent"调试
references/debug.md
"工具测试失败"、"回调测试出问题了"调试
references/debug.md
如果意图不明确，询问："您是想要构建/创建评估、运行评估，还是调试故障？"

Step Tracking

步骤跟踪

CRITICAL: Before starting any multi-step workflow (Build, Run, or Debug), you MUST create a 
todo.md
checklist in the project workspace.
As you complete each step in the workflow, update the checklist to check it off (
[ ]
to 
[x]
). Do not proceed to the next step until the current one is checked off. This ensures no steps are missed during complex builds or debug sessions.
关键注意事项： 在启动任何多步骤工作流（构建、运行或调试）之前，必须在项目工作区中创建一个
todo.md
检查清单。
完成工作流中的每个步骤后，更新检查清单将其标记为已完成（从
[ ]
改为
[x]
）。在当前步骤标记完成之前，不要进行下一步。这可确保在复杂构建或调试会话中不会遗漏任何步骤。

Before Starting

开始之前

Check memory for project-specific context (app ID, variable handling rules, audio scoring workarounds). If not available, ask the user.
检查内存中是否有项目特定上下文（应用ID、变量处理规则、音频评分变通方案）。如果没有，请询问用户。