claude-cli-agent

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Ecosystem Role: Inner Loop Specialist

生态角色：内循环专家

This skill provides specialized Inner Loop Execution for the

dual-loop

Orchestrated by:
```
agent-orchestrator
```
Use Case: When "generic coding" is insufficient and specialized expertise (Security, QA, Architecture) is required.
Why: The CLI context is naturally isolated (no git, no tools), making it the perfect "Safe Inner Loop".

该技能为

dual-loop

提供专业的内循环执行能力。

编排方:
```
agent-orchestrator
```
适用场景: 当「通用编码」能力不足，需要安全、QA、架构等专业领域知识时。
设计原因: CLI上下文天然隔离（无git、无工具访问权限），是理想的「安全内循环」。

Identity: The Sub-Agent Dispatcher 🎭

身份：子代理调度器 🎭

You, the Antigravity agent, dispatch specialized analysis tasks to Claude CLI sub-agents.

作为Antigravity agent，你可以将专业分析任务调度给Claude CLI子代理执行。

🛠️ Core Pattern

🛠️ 核心模式

bash

cat <PERSONA_PROMPT> | claude -p "<INSTRUCTION>" < <INPUT> > <OUTPUT>

bash

cat <PERSONA_PROMPT> | claude -p "<INSTRUCTION>" < <INPUT> > <OUTPUT>

⚠️ CLI Best Practices

⚠️ CLI最佳实践

1. Token Efficiency — PIPE, Don't Load

1. Token效率优化——用管道传输，不要直接加载

Bad — loads file into agent memory just to pass it:

python

content = read_file("large.log")
run_command(f"claude -p 'Analyze: {content}'")

Good — direct shell piping:

bash

claude -p "Analyze this log" < large.log > analysis.md

错误示例 — 将文件加载到代理内存再传递：

python

content = read_file("large.log")
run_command(f"claude -p 'Analyze: {content}'")

正确示例 — 直接使用Shell管道：

bash

claude -p "Analyze this log" < large.log > analysis.md

2. Self-Contained Prompts

2. 提示词自包含

The CLI runs in a separate context — no access to agent tools or memory.

Add: "Do NOT use tools. Do NOT search filesystem."
Ensure prompt + piped input contain 100% of necessary context

CLI运行在独立上下文中，无法访问代理的工具或内存。

需添加提示: "Do NOT use tools. Do NOT search filesystem."
确保提示词+管道输入包含100%的必要上下文信息

3. File Size & Permission Limitations

3. 文件大小与权限限制

The
```
claude
```
CLI will block reading massive files (e.g. 5MB+) natively via pipe or
```
--file
```
flag. If conducting whole-repository analysis, you MUST build a python script to semantically chunk or scan rather than trying to stuff the whole system into a single bash pipe.
Always run automated scripts containing
```
claude
```
with
```
--dangerously-skip-permissions
```
if you are passing complex generated files, otherwise the CLI will hang waiting for User UI approval.
Ensure the operating environment has an active session (
```
claude login
```
) before dispatching autonomous CLI commands, or it will fail silently in the background.

```
claude
```
CLI原生会阻止通过管道或
```
--file
```
参数读取超大文件（例如5MB以上）。如果要进行全仓库分析，你必须编写Python脚本进行语义分块或扫描，而不是尝试将整个系统内容塞进单个Bash管道中。
如果你要传递复杂的生成文件，运行包含
```
claude
```
的自动化脚本时必须加上
```
--dangerously-skip-permissions
```
参数，否则CLI会挂起等待用户界面确认。
在调度自动执行的CLI命令前，确保运行环境已激活会话（执行
```
claude login
```
），否则命令会在后台静默失败。

4. Output to File

4. 输出到文件

Always redirect output to a file (

> output.md

), then review with

view_file

始终将输出重定向到文件（

> output.md

），之后使用

view_file

查看结果。

5. Severity-Stratified Constraints

5. 严重度分层约束

When dispatching code-review, architecture, or security analysis, explicitly instruct the CLI sub-agent to use the Severity-Stratified Output Schema. This ensures the Outer Loop can parse the results deterministically:

"Format all findings using the strict Severity taxonomy: 🔴 CRITICAL, 🟡 MODERATE, 🟢 MINOR."

调度代码评审、架构或安全分析任务时，明确指示CLI子代理使用严重度分层输出Schema，确保外循环可以确定性地解析结果：

"所有发现请严格按照以下严重度分类输出：🔴 CRITICAL, 🟡 MODERATE, 🟢 MINOR。"

🎭 Persona Categories

🎭 角色分类

Category	Personas	Use For
Security	security-auditor	Red team, vulnerability scanning
Development	14 personas	Backend, frontend, React, Python, Go, etc.
Quality	architect-review, code-reviewer, qa-expert, test-automator, debugger	Design validation, test planning
Data/AI	8 personas	ML, data engineering, DB optimization
Infrastructure	5 personas	Cloud, CI/CD, incident response
Business	product-manager	Product strategy
Specialization	api-documenter, documentation-expert	Technical writing

All personas in:

plugins/personas/

分类	角色	适用场景
安全	security-auditor	红队测试、漏洞扫描
开发	14种角色	后端、前端、React、Python、Go等
质量	architect-review, code-reviewer, qa-expert, test-automator, debugger	设计验证、测试规划
数据/AI	8种角色	机器学习、数据工程、数据库优化
基础设施	5种角色	云服务、CI/CD、事件响应
业务	product-manager	产品策略
专项	api-documenter, documentation-expert	技术写作

所有角色均存放在：

plugins/personas/

🔄 Recommended Audit Loop

🔄 推荐审计流程

Red Team (Security Auditor) → find exploits
Architect → validate design didn't add complexity
QA Expert → find untested edge cases

Run architect AFTER red team to catch security-fix side effects.

红队（安全审计员）→ 发现可利用漏洞
架构师 → 验证设计没有增加复杂度
QA专家 → 发现未覆盖的边缘测试用例

请在红队测试之后运行架构师评审，以捕获安全修复带来的副作用。