Pre-action boundary checking — validates agent tool calls against declared capabilities and task contracts
中文简介正在准备中。当前显示原文说明。
Show help information for commands and system
中文简介正在准备中。当前显示原文说明。
安装前可以查看交给 AI 代理的指令内容。
Show help information for commands, agents, and system rules.
--agents, -a List all available agents
--commands, -c List all commands (same as 'lists')
--rules, -r List all rules
AI Agent System - Help
Usage: {command} [arguments] [options]
Quick Start:
lists Show all available commands
status Show system status
help <command> Get help for a specific command
Common Commands:
/update-docs Sync documentation with project
/update-external Update external agents
/audit-agents Check agent dependencies
/create-agent Create a new agent
Use "lists" to see all available commands.
Use "help <command>" for detailed help.
help /update-docs
Command: /update-docs
Description:
Sync documentation with project structure. Ensures all
documentation accurately reflects the current project state.
Usage:
/update-docs
/update-docs --check
/update-docs --target <path>
Options:
--check, -c Check only, don't modify
--verbose, -v Show detailed changes
--target, -t Specific target to update
Examples:
/update-docs # Update all documentation
/update-docs --check # Check for issues
/update-docs --target agents # Update agents only
help --agents
Available Agents:
Orchestrator:
secretary Manages manager agents
Manager:
mgr-creator Creates new agents
mgr-updater Updates external sources and docs
mgr-supplier Validates dependencies
SW Engineer:
lang-golang-expert Go development (Effective Go)
lang-python-expert Python development (PEP 8)
lang-rust-expert Rust development (API Guidelines)
lang-kotlin-expert Kotlin development (JetBrains)
lang-typescript-expert TypeScript development (Google)
fe-vercel-agent React/Next.js (Vercel)
Backend Engineer:
be-fastapi-expert FastAPI (Python async)
be-springboot-expert Spring Boot (Java)
be-go-backend-expert Go backend (Uber style)
Infra Engineer:
infra-docker-expert Docker containerization
infra-aws-expert AWS architecture
Total: 15 agents
help --rules
Global Rules:
MUST (Never violate):
R000 Language Policy Korean I/O, English files
R001 Safety Rules Prohibited actions
R002 Permission Rules Tool tiers, file access
R006 Agent Design Structure, separation
SHOULD (Strongly recommended):
R003 Interaction Rules Response format
R004 Error Handling Error levels, recovery
R007 Agent Identification Display agent in responses
R008 Tool Identification Display agent when using tools
MAY (Optional):
R005 Optimization Efficiency guidelines
R009 Parallel Execution Max 4 parallel instances
Total: 10 rules
还没有评价。欢迎分享你的使用感受。
无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
Pre-action boundary checking — validates agent tool calls against declared capabilities and task contracts
中文简介正在准备中。当前显示原文说明。
Auto-detect project context and optimize harness — deactivate unused agents/skills, suggest missing experts, generate project profile
中文简介正在准备中。当前显示原文说明。
Adversarial code review using attacker mindset — trust boundary, attack surface, business logic, and defense evaluation
中文简介正在准备中。当前显示原文说明。
Apache Airflow best practices for DAG authoring, testing, and production deployment
中文简介正在准备中。当前显示原文说明。
Alembic migration patterns for naming conventions, safety checks, expand-contract, env.py configuration, and CI integration
中文简介正在准备中。当前显示原文说明。
Pre-routing ambiguity analysis — scores request clarity and asks clarifying questions when needed (inspired by ouroboros)
中文简介正在准备中。当前显示原文说明。