npx agentspec-cli install 5bb1066d-7604-4f54-8ace-8e9ad6233c8b --file-type agents-md→ .agents/
# AGENTS.md - Guide for AI Coding Assistants This document provides development guidelines for AI-powered coding tools working on the CCoDoLo project itself. **For project features, usage, and architecture**: See [README.md](README.md) ## What This Project Is CCoDoLo is a meta-project - it provides sandboxed Docker containers for running AI coding assistants (including you). When modifying this codebase, you're improving the infrastructure that runs AI agents safely. ## Repository Structure ``` docker/ ├── ccodolo # Main shell script - container launcher and project manager ├── bin/start-agent # Container entrypoint (runs inside container) ├── docker/Dockerfile # Container image definition └── dotfiles/ # Shell configurations (zsh, bash, powerline10k) ``` Projects are stored in `~/.ccodolo/projects/<name>/` (gitignored). ## Development Guidelines ### Shell Script Validation **Always validate shell scripts before committing:** ```bash sh -n ccodolo sh -n bin/start-agent bash -n dotfiles/.bashrc zsh -n dotfiles/.zshrc ``` The `ccodolo` script must remain POSIX-compatible (#!/bin/sh) for portability. ### Docker Best Practices - Pin base image versions for reproducibility - Document new tools in README.md "Installed Tools" section - Keep the image lean - only include tools needed across multiple agents - Test builds: `docker build -t ccodolo:test -f docker/Dockerfile .` ## Testing Changes Before committing: 1. **Syntax validation**: `sh -n` on all modified shell scripts 2. **Build test**: Verify container builds successfully 3. **Runtime test**: Create test project and verify agent launches 4. **Multi-agent test**: Test with 2+ agents if touching agent selection logic ```bash # Quick test flow ./ccodolo --project test-changes --agent claude --create-new # Inside container, verify mounts and agent launch exit ./ccodolo --project test-changes --agent copilot # Test switching ``` ## Common Modifications **Adding a new agent**: Update `ccodolo` (valid agents list, mount logic, help text), update `start-agent` (launch command), document in README.md (see existing agent sections for pattern). **Adding tools**: Add to Dockerfile, document in README.md. Consider if tool is universal or agent-specific. **Shell configuration**: Edit `dotfiles/`, ensure both zsh and bash work, test history persistence, rebuild and verify. ## Key Constraints - `ccodolo` script must remain POSIX sh-compatible - Environment variables are NOT passed from host to container - Both zsh and bash must have feature parity - Project isolation is critical - test that configs don't leak between projects ## Testing Checklist for PRs - [ ] Shell scripts validated with `sh -n` - [ ] Container builds successfully - [ ] Tested with at least one agent end-to-end - [ ] README.md updated if adding features/tools - [ ] Backward compatible with existing projects
Agents.md — Guidance for AI-assisted Development
_An overview of the current repository and its related projects, generated by [AgentPrimer](https://github.com/deanthecoder/AgentPrimer)._
by AgentHub Crawler
Agents.md – Guidance for OpenAI Codex
This `Agents.md` file provides detailed instructions for OpenAI Codex and other AI agents collaborating on this repository. It includes project structure, conventions, testing, and update policies.
by AgentHub Crawler
Agent Guidelines for .agents
This repository contains agent definitions, command workflows, skill definitions, and coding rules for AI-powered coding tools. These files configure how specialized AI agents behave when assisting with development tasks.
by AgentHub Crawler
Sign in to leave a comment.
Agent Instructions
This document provides guidelines for building, testing, and contributing to the CCCL repository. It is primarily written for agentic AIs, but the information is also useful for CCCL developers.
by AgentHub Crawler