Skip to content

AGENTS.md

Latest commit

 

History

History
66 lines (47 loc) · 2.37 KB

AGENTS.md

File metadata and controls

66 lines (47 loc) · 2.37 KB

Project Instructions

Agentic Workflow Engine: Scheduling, Orchestrating, Managing Events

Domain Model

Four nouns: Event, Worker, Task, Policy. One pipeline. See .claude/rules/domain-model.md.

Tech Stack

  • Go 1.26.2 — single binary, cross-platform (darwin/linux, amd64/arm64)
  • Cobra CLI, YAML config, JSON file persistence, Docker API
  • VitePress for documentation (website/)

Build

make build    # → ./mecha binary
make test     # go test ./...
make ci       # vet + test + build

Docs Build

cd website
npm install          # first time only
npm run dev          # dev server at localhost:5173
npm run build        # production build → .vitepress/dist/

Config: website/.vitepress/config.mts Custom styles: website/.vitepress/theme/custom.css

Architecture Docs

Design decisions and rationale are in dev-docs/ (gitignored, local only).

Guidelines

  • Prefer ${CLAUDE_CODE_OAUTH_TOKEN} over ${ANTHROPIC_API_KEY} in all configs and examples
  • In documentation, always use Mermaid charts to illustrate pipelines, state machines, and architecture whenever possible
  • Always validate Mermaid syntax after writing it (use the mcp__mermaider__validate_syntax tool before committing)
  • Codex model: use gpt-5.4, not gpt-5.2-codex
  • When a Codex task requires model selection, run /codex-toolkit:preflight first to verify connectivity and discover available models
  • Save Playwright screenshots to .playwright-mcp/ or /tmp/, never the repo root
  • Save grill reports to dev-docs/grills/, not the repo root

Shared Memory

Write shared instructions to AGENTS.md. Write enforceable Go rules to .claude/rules/.

  • AGENTS.md — project-wide context, guidelines, and structure (shared across all AI tools)
  • .claude/rules/*.md — scoped, enforceable rules auto-loaded by glob pattern (Go conventions, domain model, security)
  • Never modify CLAUDE.md or GEMINI.md directly — they only import AGENTS.md

Project Structure

  • .claude/agents/ - Custom subagents for specialized tasks
  • .claude/skills/ - Claude Code skills (slash commands)
  • .claude/rules/ - Modular rules auto-loaded into context
  • .codex/skills/ - Codex CLI skills
  • .codex/prompts/ - Codex CLI custom slash commands
  • .gemini/skills/ - Gemini CLI skills
  • .gemini/commands/ - Gemini CLI custom slash commands (TOML)
  • .mcp.json - MCP server configuration