# agent-foundry

**Repository Path**: CatCakeCandle/agent-foundry

## Basic Information

- **Project Name**: agent-foundry
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2026-04-28
- **Last Updated**: 2026-04-28

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Agent Foundry

Agent Foundry is a Codex skill for designing AI coding agents, CLI assistants, tool execution systems, permission and sandbox models, MCP integrations, streaming loops, prompt caching, context compaction, and multi-agent workflows.

The repository root is the skill root. Install or publish this repository directly; no nested `docs/agent-foundry` path is required.

## Structure

- `SKILL.md`: skill entry, trigger metadata, workflow, architecture checklist
- `agents/openai.yaml`: UI metadata for skill lists and default prompt
- `references/reference-map.md`: topic routing for progressive loading
- `references/*.md`: Claude Code v2.1.88 source-analysis notes adapted into reusable architecture references

## Reference Topics

- `references/architecture-overview.md`: layers, control flow, extensibility
- `references/long-running-agent-loop.md`: long-running agent state machine, self-repair, retries, heartbeats, compaction, implementation checklist
- `references/api-and-services.md`: model API calls, retries, streaming, MCP, compaction, analytics
- `references/tool-system.md`: tool contracts, execution, file edit/search tools, sub-agent tools
- `references/permissions-and-security.md`: permission modes, classifiers, hooks, sandboxing, path and command safety
- `references/patterns-and-practices.md`: TypeScript patterns, errors, async streams, state, performance, testing

## Use

Use this skill when another project needs Claude Code-inspired architecture guidance without copying Claude Code-specific implementation choices.