ChrisTitusTech Releases 'titstack' Repository for Cross-Platform Codex and AI Development Skills

Overview of the Titstack Repository

A new open-source project named `titstack`, hosted by developer Chris Titus, provides a reusable Codex configuration and skill inventory designed for Linux, macOS, and Windows environments. The repository offers shared instructions, command rules, optional model profiles, and development skills tailored for system administration and software engineering.

According to the official documentation, the setup is compatible with both Bash-based utilities on Unix-like systems and PowerShell on Windows. The project acts as a configuration manager for an already installed and authenticated Codex CLI or desktop application, rather than providing the underlying model access or software installation binaries itself.

Installation and Configuration Mechanics

Users are instructed to clone the repository into a permanent location, as instructions, rules, profiles, and skills are linked via symbolic links directly to the checkout directory. Editing those source files immediately updates the active environment.

For Windows users, setup requires enabling Developer Mode to permit the creation of symbolic links without an administrative terminal. Platform-specific installers automatically handle file placement, while backing up existing configuration files into a designated backup directory under `~/.codex/backups/`.

The default configuration ships with opinionated settings, notably pointing to the GPT-6 Astra model, disabling approval prompts (`approval_policy = "never"`), and granting full filesystem access (`sandbox_mode = "danger-full-access"`). It also authorizes local commits and automatically trusts repositories located under the user's `~/github` directory. The documentation stresses that users should review and adjust these security boundaries to match their personal workflows before running the installation script.

Workflow Integration and Plugins

Titstack incorporates a structured lifecycle separating initial planning from pull-request readiness. Built-in workflow agents manage project documentation such as `AGENTS.md`, `SPEC.md`, `ROADMAP.md`, and `TASKS.md`, allowing developers to execute phased changes methodically.

Plugin installation is strictly opt-in via a manifest file (`codex-plugins.txt`), with the curated `superpowers@openai-curated` plugin selected initially to assist with planning, testing, and debugging. Additional integrations, such as the native `photocraft-cli` MCP server for document and image creation, are supported through optional setup steps.

Practical Implications and Verification

Developers evaluating the repository should carefully consider the following operational details:

• Confirmed Features: Shared instruction sets, symbolic-link-based updates, automated trust-entry generation for local Git repositories, and support for both local `llama.cpp` endpoints and default cloud models.
• Unknowns & Customizations: Third-party plugin behaviors, unverified public adoption metrics, and the specific behavior of custom model integrations not covered by default configurations.
• Security Considerations: Default settings prioritize high autonomy and broad filesystem access. Security-conscious users are strongly advised to modify `approval_policy` and `sandbox_mode` settings prior to deployment.

Readers can examine the complete source code, installation scripts, and documentation directly on the [GitHub Repository](https://github.com/ChrisTitusTech/titstack).