Gav Verma

This commit **refactors** the **app-based skills** subsystem by **removing its direct dependency on the `mcp` component**. This change streamlines the architecture for app-based skills, making them more self-contained and reducing unnecessary coupling. The primary impact is improved modularity and potentially reduced overhead for skill development and deployment, as `app-based skills` no longer require `mcp` to function.

This commit implements a **configuration adjustment** to the **plugin marketplace** by adding an empty `products` array within the `policy` section of numerous plugins in `marketplace.json`. This change effectively **filters out** these specific plugins, preventing them from being returned or displayed to users. The work ensures that certain plugins are no longer visible, directly impacting the **availability and presentation of plugins** within the system.

This commit **refactors** the **plugin definition specification** by changing the `interface.defaultPrompt` field from a single string to an array of up to three strings. This **enhances the plugin-creator skill** by allowing multiple starter prompts for plugins, improving user guidance and flexibility. The change involves updating the `plugin-json-spec.md` with new truncation rules and a reduced per-entry character limit, and modifying the `create_basic_plugin.py` script to scaffold new plugins accordingly. Additionally, **over a hundred existing `plugin.json` files** across various plugins were updated to conform to this new array format, with their prompts normalized and shortened.

This commit introduces **enhanced support for marketplace policies** by **refactoring** the plugin policy schema. It consolidates disparate `installPolicy` and `authPolicy` fields into a new, **nested `policy` object** within **plugin marketplace entries**, which now also accommodates `policy.products`. This change standardizes and improves the extensibility of how **plugin installation and authentication policies** are defined and managed. The update includes comprehensive **documentation updates** for the `plugin-creator` skill and its JSON specification, along with modifications to the `create_basic_plugin.py` script to generate compliant plugin entries. This ensures all new and existing plugins can leverage a more robust and organized policy framework.

This commit **introduces a new `interface.displayName` field** to the `marketplace.json` configuration, significantly enhancing the metadata for plugins. The **`plugin-creator` skill** is updated to **support this new capability**, with the `create_basic_plugin.py` script now handling default values, validating the structure via `validate_marketplace_interface`, and integrating it into the marketplace building logic. This **feature implementation** provides a user-friendly display name for plugins, improving their discoverability and presentation within the system. **Documentation** for the `plugin-creator` skill (`SKILL.md`) and the `plugin-json-spec.md` has also been updated to reflect this important addition.

This commit **refines agent-facing skill descriptions** across numerous **plugins**, including `game-studio`, `gmail`, `google-calendar`, `google-docs`, `linear`, `netlify`, `outlook-calendar`, `slack`, and others. This **documentation refinement** involved rephrasing, clarifying, and making descriptions more concise or specific for individual skills like `game-playtest`, `google-sheets`, and `outlook-email`. The **maintenance** work aims to improve the clarity and interpretability of skill capabilities for AI agents, enhancing their ability to accurately select and utilize the appropriate tools for various tasks. These updates span a wide array of functionalities, from game development to productivity and communication tools, ultimately improving the overall **agent skill routing and utilization** across the system.

This commit **updates the categorization of plugins** within the **marketplace configuration**. It refines the existing category structure by replacing the general 'Productivity' category with more specific options such as 'Lifestyle', 'Research', 'Design', and 'Coding' in the `.agents/plugins/marketplace.json` file. This **configuration update** is a **chore** that improves the organization and discoverability of plugins for users browsing the marketplace.

This commit introduces **support for `marketplace.json`** within the **`plugin-creator` skill**, enabling the generation and management of marketplace entries for plugins. It implements new functionality in `create_basic_plugin.py` to create or update these entries, including new command-line arguments and helper functions like `build_marketplace_entry`. This **new feature** is accompanied by comprehensive **documentation updates**, including a new specification for `marketplace.json` in `plugin-json-spec.md` and workflow details in `SKILL.md`, ensuring users can leverage this new capability. The agent's default prompt in `openai.yaml` has also been updated to reflect the new marketplace entry scaffolding.

This commit introduces a **new feature** by adding a **Google Calendar daily brief skill** to the `google-calendar` plugin. This capability allows the system to process a user's calendar events and generate a comprehensive, structured daily summary in Markdown, detailing scheduled events, free time, and potential conflicts. The implementation includes a new Python script, `render_day_brief.py`, responsible for event processing and rendering, alongside updated plugin metadata and dedicated OpenAI agent configuration for the new skill. This significantly enhances the **Google Calendar plugin's** utility by providing users with an automated, intelligent overview of their day.

This commit **introduces the initial scaffolding for the plugin marketplace system** by adding the `.agents/plugins/marketplace.json` configuration file. This **new feature** defines an `openai-curated` marketplace, which includes a `linear` plugin for testing purposes. It establishes the foundational manifest for future plugin discovery and management within the system, laying the groundwork for how plugins will be sourced and configured.

This commit performs a **documentation update** by refining the `README.md` file. It specifically **updates the repository's title** and replaces generic placeholder text with a more accurate and informative description of the project's contents. This **maintenance** task enhances the initial understanding for anyone encountering the repository, improving overall clarity and user experience for the entire project.

This commit **introduces a new Linear plugin**, providing a **new capability** for integration with the Linear platform. This includes all essential components such as the core plugin configuration (`plugin.json`), server-side settings (`.mcp.json`), and the Apache 2.0 license for the **Linear skill**. Furthermore, it adds comprehensive Markdown documentation (`SKILL.md`) and an OpenAI agent configuration (`openai.yaml`), enabling the system to interact with Linear through this new module.

This commit **fixes incorrect filenames** used for `mcpServers` and `apps` configurations within the **`plugin-creator` skill**. It **updates the `build_plugin_json` function** in `create_basic_plugin.py` to generate plugin configurations with the correct **dot-prefixed filenames** (`.mcp.json` and `.app.json`). Concurrently, the **plugin JSON specification documentation** (`plugin-json-spec.md`) has been corrected to reflect these accurate filenames. This **bug fix** ensures consistency between generated plugins and their documentation, preventing potential issues related to file naming conventions.

This commit introduces a **new capability** by creating a `plugin-creator` skill, designed to streamline the development of new plugins. It provides a **scaffolding tool** via the `create_basic_plugin.py` script, which handles directory creation, plugin name normalization, and the generation of a placeholder `plugin.json` file. The commit also includes comprehensive **documentation** for the skill itself in `SKILL.md` and a detailed reference for the plugin JSON specification in `plugin-json-spec.md`, along with an OpenAI agent interface configuration. This new skill significantly simplifies the initial setup phase for plugin development within the system.

This commit introduces a **new feature** to the **skill loading mechanism**, enabling the system to read **personal skills** from the user's home directory at `$HOME/.agents/skills`. This expands the existing capability, which previously only loaded skills from repository-specific `.agents/skills` directories, without affecting system skill locations. The changes in `codex-rs/core/src/skills/loader.rs` enhance user customization by allowing global skill definitions. This update maintains backward compatibility with `~/.codex/skills` and provides a more flexible way for users to manage their personal skill sets.

This commit **fixes a critical bug** in the **system skills management** where bundled skills were not correctly replacing user skills due to an outdated marker. It addresses this by **recursively hashing embedded system skills**, including nested directories and file contents, to ensure the `.codex-system-skills.marker` updates properly and triggers a reinstall. A new build Cargo hook in `codex-rs/core/build.rs` was also added to force a rebuild when `src/skills/assets/samples/*` changes, guaranteeing that embedded skill updates are always reflected. This ensures **correct and timely updates of system skills** for users, preventing stale skill sets and improving the reliability of skill deployment.

This commit **syncs system skills** from an external repository, introducing new configurations and tooling primarily for OpenAI-related integrations within the **skills subsystem**. It **adds `openai.yaml` configuration files** for both the `skill-creator` and `skill-installer` agents, defining their display names and descriptions. A new **Python script (`generate_openai_yaml.py`)** is included to automate the creation of these YAML files, alongside another script (`list-skills.py`) to facilitate skill discovery and management. This **feature addition** streamlines skill configuration and installation, enhancing the overall capability for integrating and managing skills, and includes new documentation for the `openai.yaml` structure.

This commit **synchronizes the system skills** within `codex-rs` with the latest updates from the public `openai/skills` repository. It **updates documentation** for skill creation and installation, specifically adding details on new agent metadata, `openai.yaml` generation, and usage of experimental skills. Furthermore, the `init_skill.py` script gains **new functionality** to generate `agents/openai.yaml` and accept interface overrides, streamlining the skill initialization process. This **maintenance and feature update** ensures `codex-rs` users have access to the most current skill development practices and tools.

This commit introduces **support for loading skills from `.agents/skills/` directories**, providing a standardized and more convenient location for sharing skills across agents. It **enhances the skill discovery logic** within the `codex-rs/core/src/skills/loader.rs` module and updates the `codex-rs/core/src/skills/manager.rs` to utilize this new path. This **new capability** simplifies skill management by allowing skills to be placed in a central, repository-level location, addressing previous difficulties with sharing and duplication. While existing `.codex/skills/` paths will still function, this change primarily affects **REPO-scoped skill loading** and sets the stage for the deprecation of the older path.

This commit performs a **documentation update** for the **`app-server`** module. Specifically, it revises the example JSON response for the **`skills/list` protocol** within the `README.md` file. The update incorporates **new interface fields** into the example, ensuring the documentation accurately reflects the latest API specification. This **maintenance** task helps developers correctly understand and integrate with the `skills/list` protocol by providing up-to-date response examples.