Production agent architecture
Skill xuanhieu2611/build-your-own-agents-skill/skills/production-agent-architecture
Build AI agents using the same architecture patterns as Claude Code. Skill + 6 examples + runnable Python code. No framework.
npx -y skills add xuanhieu2611/build-your-own-agents-skill --skill production-agent-architectureAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 24 stars24 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Design production-ready AI agents using a loop-based architecture with explicit tools, context, memory, permissions, observability, evaluation, and rollout planning.
SKILL.md
2.0 KB, as published. Nobody here has run it
Production Agent Architecture
When To Use This Skill
Use this skill when the task is to:
- design a new AI agent
- turn a chatbot idea into a real agent system
- hand off an agent build to developers or another AI coding agent
- adapt coding-agent architecture patterns to another domain
Core Rule
Do not design the system as a one-shot chatbot unless the user explicitly asks for that.
Default to a loop-based runtime with:
- controller loop
- context builder
- tool registry
- tool executor
- session state
- permissions and approvals
- observability
- evaluation
- reliability rules
- rollout phases
Design Workflow
- define the user goal and success condition
- define the controller loop
- define 2 to 5 domain tools
- define context inputs
- define session state
- define permissions and approval rules
- define observability minimums
- define evaluation plan
- define reliability and retry rules
- define rollout phases
Required Output
Always produce all of the following:
# Agent Build Spec
## Domain
## User goal
## Success condition
## Core workflow
## Controller loop
## Context inputs
## Tools
## Session state
## Permissions and approvals
## Stop conditions
## Observability
## Evaluation
## Reliability and failure handling
## Rollout phases
## Future extensions
After the spec, also produce:
- controller loop pseudocode
- tool contract table with input, output, side effects, timeout, and permission level
- session state schema
- approval matrix
- failure and retry strategy
- minimum logs and metrics
Guardrails
- separate tool schemas from executors
- treat permissions as runtime policy, not prompt text
- make approval flows explicit
- do not pretend a vague idea is production-ready
- explain the design plainly if the user is learning