tcaitoolkit: Instructions file for Codex

AGENTS.md for Thinking Craftsman AI Toolkit

This file is for AI Agents working on this project. README.md is for Humans working on this project.

Overview

This project is a set of utilities to help human software developers adopt AI Coding Agents in their existing or new projects. This is developed by Nitin Bhide

Project Structure

• ./docs/ - documents folder
• ./docs/agentskills/ - folder for specification and design documents of various Agent Skills
• ./docs/docindex.md - index of specification and design documents.
• ./agentic_cc/ - source of python cookie cutter to generate the folder structure for a new project. Use this when you want to start a new project with use of AI Coding Agents from the day one.
• ./agentskills/ - Useful agent skills like grill-me, repo-nav, agent-audit, thinking-craftsman-skill for coding and code reviews etc.

Testing Bash Scripts from Windows with WSL

• Windows checkouts may give shell scripts CRLF line endings, which can make WSL Bash fail with errors such as $'\r': command not found or a syntax error near in\r. Do not change tracked line endings just to test. Normalize temporary copies with sed 's/\r$//'.

• If a script sources helpers or reads other files using paths relative to itself, copy the script and its dependencies to a temporary directory while preserving their relative directory structure. Invoke temporary scripts with bash "$tmp_dir/path/to/script.sh"; the copies may not have executable permissions.

• For syntax checks, run bash -n on each normalized copy. For behavior checks, run the normalized script with bash against a temporary fixture and clean up the fixture afterward.

• When launching a multiline Bash test from PowerShell, pass it to WSL as one argument. Encode the here-string before invoking wsl.exe to avoid PowerShell native-argument splitting or expansion:

  $encoded = [Convert]::ToBase64String([Text.Encoding]::UTF8.GetBytes($bashCmd))
  wsl.exe bash -lc "printf '%s' '$encoded' | base64 -d | bash"