repo-nexus 0.7.2 → 0.7.4

package/CHANGELOG.md

@@ -7,7 +7,15 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
-## [0.6.16] - 2026-10-14
+### Added
+- **Two-Tier Skills Architecture**: Clean separation between core `rnex` skills (`toolkit/skills/*.md` -> `.rnex/skills/*.md`) and plugin-owned skills (`toolkit/plugins/<plugin>/skills/*.md` -> `.rnex/plugins/<plugin>/skills/*.md`).
+- **On-the-Fly User Skill Generator**: Generated in active AI provider directories (`.cursor/rules/`, `.claude/skills/`, `.gemini/skills/`, `.windsurf/rules/`) with mandatory `rnex-` prefix (`rnex-<name>.md`). User skills contain zero inline heuristics, linking solely to the target agentic skill protocol in `.rnex/`.
+- **Plugin Contract Helper API**: Added `rnex_register_user_skill` and `rnex_unregister_user_skill` helper functions exposed in hook subshell environments for plugin lifecycle management.
+- **Plugin Lifecycle Hooks**: Added `before_plugin_disable` hook dispatched by `rnex plugin disable <name>` and plugin pruning in `rnex sync` to unregister plugin-owned User Skills.
+
+### Removed
+- **Legacy Prompts & Workflows**: Completely removed `toolkit/prompts/`, `toolkit/workflows/`, and `.rnex/workflows/`.
+- **Copilot Local Prompt Harness**: Removed `toolkit/plugins/copilot/`, `--copilot` CLI flag, and prompt-mirroring harness logic in favor of native universal `AGENTS.md` reading.
 
 ### Removed
 - **`rnex fix` and `rnex status --fix`**: Removed in favor of `rnex sync` as the sole reconciliation command.

package/README.md

@@ -157,11 +157,9 @@ repo-nexus/
 │   ├── test_cli.sh                 # Main test runner
 │   ├── test_helper.sh              # Shared test harness and utilities
 │   └── suites/                     # Modular test suites
-├── toolkit/                        # Built-in plugins, prompts, skills, and workflows
+├── toolkit/                        # Built-in plugins and core agentic skills
 │   ├── plugins/                    # Scoped workspace plugins
-│   ├── prompts/                    # Standardized AI operational prompts
-│   ├── skills/                     # Workspace skills
-│   └── workflows/                  # Workspace workflows
+│   └── skills/                     # Core workspace agentic skills
 └── README.md
 ```
 

package/docs/CLI.md

@@ -35,9 +35,9 @@ rnex sync [-y|--yes] [-q|--quiet]
 ### Reconciliation Actions
 1. **Config Version Upgrade**: Detects schema drift and non-destructively upgrades `rnex.yaml` if older than the installed CLI version.
 2. **Git Ignore & Search Indexing**: Reconciles `.gitignore` (ignoring `repos/*` and `.local.rnex.yaml`) and `.ignore` (un-ignoring `repos/*` for search tools).
-3. **Plugin Sync**: Prunes disabled plugins and deterministically copies enabled plugin assets, rules, workflows, templates, and executable lifecycle hooks into `.rnex/plugins/<plugin-name>/`.
+3. **Plugin Sync**: Prunes disabled plugins and deterministically copies enabled plugin assets, rules, skills, templates, and executable lifecycle hooks into `.rnex/plugins/<plugin-name>/`.
 4. **Compiled AI Instructions**: Compiles `.rnex/ai-instructions.md` by combining the base template with active configuration and all enabled plugin `instructions.md` fragments. Top-level provider files (`AGENTS.md`, `CLAUDE.md`, etc.) receive only a single strict directive pointing to `.rnex/ai-instructions.md`.
-5. **Prompt Sync**: Copies all operational prompts matching `rnex-*` from `toolkit/prompts/` into configured AI assistant directories (Claude, Gemini, Cursor, Windsurf, Copilot) and workspace prompt directories.
+5. **Skill Reconciliation & User Skill Generation**: Copies core agentic skills from `toolkit/skills/` to `.rnex/skills/`, and generates lightweight, zero-logic User Skills (prefixed with `rnex-`) in configured AI assistant directories (`.cursor/rules/`, `.claude/skills/`, `.gemini/skills/`, `.windsurf/rules/`) pointing directly to `.rnex/skills/` or plugin skill directories.
 6. **Harness Sync**: Reconciles root `.rnex/` and member repository harnesses (`repos/<name>/.rnex/`) for all enabled member repositories where `rnex_dir: true`.
 7. **Git Hooks**: Installs or reconciles workspace Git hooks in `.rnex/hooks` when `git_hooks: true`.
 8. **Multi-Root Workspace**: Synchronizes `.code-workspace` folder mappings when `code_workspace: true`.
@@ -66,13 +66,14 @@ toolkit/plugins/<plugin-name>/
 ├── hooks/                       # POSIX /bin/sh hook scripts (chmod +x)
 │   ├── after_sync.sh            # Executes after workspace reconciliation
 │   ├── after_init.sh            # Executes after workspace initialization
+│   ├── before_plugin_disable.sh # Executes before plugin disable/removal
 │   └── before_remove.sh         # Executes before member repo removal
 ├── rules/                       # AI assistant behavioral rules
-├── workflows/                   # Standardized agent workflow procedures
+├── skills/                      # Plugin-owned agentic skills (e.g. wiki-ingest.md)
 └── templates/                   # Default files and scaffold templates
 ```
 
-### Environment Context Passed to Hooks
+### Environment Context & Helper Functions
 Every lifecycle hook receives the following environment variables:
 
 | Variable | Description |
@@ -80,10 +81,15 @@ Every lifecycle hook receives the following environment variables:
 | `RNEX_ROOT` | Absolute path to the Repo Nexus workspace root directory. |
 | `RNEX_CONFIG_MERGED` | Path to the active workspace configuration (`rnex.yaml`). |
 | `RNEX_PLUGIN_DIR` | Path to the scoped plugin directory (`.rnex/plugins/<plugin-name>`). |
+| `RNEX_PROVIDER_SKILL_DIRS` | Space-separated list of active provider skill directories. |
 | `RNEX_CMD` | The command being executed (e.g., `sync`, `init`, `add`). |
 | `RNEX_ARGS` | Arguments passed to the command. |
 | `RNEX_QUIET` | Set to `1` when `--quiet` was passed, `0` otherwise. |
 
+Hooks have access to POSIX helper functions for managing User Skill pointers:
+- `rnex_register_user_skill <skill_name> <target_path_in_rnex>`: Generates a zero-logic user skill pointer (`rnex-<skill_name>.md`) in active provider directories linking to the target skill protocol.
+- `rnex_unregister_user_skill <skill_name>`: Removes the `rnex-<skill_name>.md` user skill pointer from all active provider directories.
+
 ### Karpathy LLM Wiki Plugin
 The `karpathy-llm` plugin uses lifecycle hooks to manage its domain directories:
 - **Default curated knowledge base (`wiki_dir`)**: `.rnex/wiki`

package/docs/ai-instructions.sample.md

@@ -7,7 +7,7 @@ This workspace operates as a **Repo Nexus Virtual Meta-Repo** uniting independen
 ### Mandatory Session Startup
 Execute in sequence before task execution:
 1. **Read Configuration First (Highest Priority)**: Read `.local.rnex.yaml` (if present) and `rnex.yaml` to identify active member repositories (`repos/`) and enabled plugins.
-2. **Plugin & Skill Routing**: Inspect active plugin instructions under `.rnex/plugins/<plugin-name>/`, and tools under `.rnex/skills/` and `.rnex/workflows/` when relevant.
+2. **Plugin & Skill Routing**: Inspect active plugin instructions under `.rnex/plugins/<plugin-name>/`, and core agentic skills under `.rnex/skills/` when relevant.
 3. **Targeted Member Repository Access Only**: Navigate directly to specific files in `repos/<name>/` relevant to the user request. Avoid blind repository-wide scans or unbounded searches across `repos/`.
 
 ## Subagents

package/docs/cli/index.md

@@ -47,19 +47,19 @@ rnex [global-options] <command> [command-options] [arguments]
 
 ---
 
-## Standardized Prompts Catalog
+## Agentic Skills Catalog
 
-Repo Nexus provides a standardized suite of operational prompt templates in `toolkit/prompts/` that sync during `rnex sync` or `rnex init`.
+Repo Nexus provides a standardized suite of agentic skills in `toolkit/skills/` (core) and `toolkit/plugins/<plugin>/skills/` (plugins).
 
-AI assistants navigate workflows via the catalog index at `.rnex/prompts/index.md`:
+AI assistants navigate skills via `AGENTS.md` and direct skill protocol paths:
 
-| Prompt File | Purpose |
-|:---|:---|
-| `index.md` | Master catalog indexing all available workflows and prompts. |
-| `rnex-cross-repo-feature.md` | Structured workflow for implementing cross-cutting features across multiple repos. |
-| `rnex-workspace-audit.md` | Health, git dirty status, branch tracking, and configuration audit workflow. |
-| `rnex-wiki-ingest.md` | Autonomous ingestion of source docs from `/raw/` into Karpathy LLM Wiki pages. |
-| `rnex-wiki-lint.md` | Link validation, backlink (`mentioned_in`) recalculation, and index rebuilding. |
+| Skill | Category | Purpose | Protocol Path |
+|:---|:---|:---|:---|
+| `clone-sync` | Core | Automated clone, setup, and synchronization of member repositories. | `.rnex/skills/clone-sync.md` |
+| `batch-exec` | Core | Cross-repository batch commands, parallel execution, and verification. | `.rnex/skills/batch-exec.md` |
+| `workspace-status` | Core | Workspace health check, git dirty status, branch tracking, and configuration audit. | `.rnex/skills/workspace-status.md` |
+| `wiki-ingest` | Plugin (`karpathy-llm`) | Autonomous ingestion of source docs into atomic Karpathy LLM Wiki pages. | `.rnex/plugins/karpathy-llm/skills/wiki-ingest.md` |
+| `wiki-lint` | Plugin (`karpathy-llm`) | Link validation, backlink recalculation, and index rebuilding. | `.rnex/plugins/karpathy-llm/skills/wiki-lint.md` |
 
 ---
 

package/docs/cli/workspace.md

@@ -26,11 +26,10 @@ rnex init [options] [directory]
 | `--code-workspace` | | boolean \| string | `false` | Enable VS Code multi-root workspace file. Pass string for custom name. |
 | `--hooks` | | boolean | `true` (if git) | Force-enable automated Git hooks setup during initialization. |
 | `--no-hooks` | | boolean | `false` | Skip Git hooks configuration during initialization. |
-| `--claude` | | boolean | `false` | Enable Claude Code plugin (mirrors prompts to `.claude/commands/`). |
-| `--copilot` | | boolean | `false` | Enable GitHub Copilot plugin (mirrors prompts to `.github/prompts/`). |
-| `--cursor` | | boolean | `false` | Enable Cursor plugin (mirrors prompts to `.cursor/prompts/`). |
-| `--gemini` | | boolean | `false` | Enable Google Gemini plugin (mirrors prompts to `.gemini/prompts/`). |
-| `--windsurf` | | boolean | `false` | Enable Windsurf plugin (mirrors prompts to `.windsurf/prompts/`). |
+| `--claude` | | boolean | `false` | Enable Claude Code plugin (maintains `CLAUDE.md`, generates skills in `.claude/skills/`). |
+| `--cursor` | | boolean | `false` | Enable Cursor plugin (maintains `.cursorrules`, generates skills in `.cursor/rules/`). |
+| `--gemini` | | boolean | `false` | Enable Google Gemini plugin (maintains `GEMINI.md`, generates skills in `.gemini/skills/`). |
+| `--windsurf` | | boolean | `false` | Enable Windsurf plugin (maintains `.windsurfrules`, generates skills in `.windsurf/rules/`). |
 
 ### Examples
 ```bash
@@ -49,7 +48,7 @@ Performs comprehensive diagnostic and health checks across the workspace.
 
 ## `rnex sync`
 
-Reconciles workspace configuration, prompts, plugins, `.rnex/` harness directories, and compiles `.rnex/ai-instructions.md`. `rnex sync` is the sole reconciliation command in Repo Nexus (`rnex fix` and `rnex status --fix` are removed).
+Reconciles workspace configuration, skills, plugins, `.rnex/` harness directories, and compiles `.rnex/ai-instructions.md`. `rnex sync` is the sole reconciliation command in Repo Nexus (`rnex fix` and `rnex status --fix` are removed).
 
 ### Syntax
 ```sh
@@ -66,7 +65,7 @@ rnex sync [options]
 1. Upgrades configuration schema non-destructively if an older version is detected.
 2. Re-verifies `.gitignore` rules for `repos/` and local configs, and ensures `.ignore` contains `!repos/*`.
 3. Compiles `.rnex/ai-instructions.md` by combining the base template with active configuration and all enabled plugin `instructions.md` fragments. Writes single strict directive to top-level provider files (`AGENTS.md`, `CLAUDE.md`, etc.).
-4. Syncs operational prompts matching `rnex-*` from `toolkit/prompts/` to workspace and configured AI assistant prompt directories.
+4. Syncs core agentic skills from `toolkit/skills/` to `.rnex/skills/` and generates zero-logic User Skills (`rnex-*.md`) in active AI provider directories.
 5. Re-syncs enabled plugin assets into `.rnex/plugins/` and member `repos/<name>/.rnex/`.
 6. Dispatches plugin lifecycle hooks (`before_sync`, `after_sync`).
 7. Reconciles automated Git hooks in `.rnex/hooks/` against `git_hooks` config.

package/docs/examples/ai-workflows.md

@@ -7,34 +7,35 @@ description: Practical walkthrough of executing cross-cutting features, workspac
 
 One of the greatest challenges with AI coding assistants in multi-repo setups is context isolation: agents operating inside one repository cannot easily coordinate changes across upstream or downstream services.
 
-Repo Nexus equips AI assistants with standardized cross-repo skills and workflows located in `.rnex/skills/` and `.rnex/workflows/`, routed directly through universal `AGENTS.md`.
+Repo Nexus equips AI assistants with standardized cross-repo agentic skills located in `.rnex/skills/` and active plugin directories (`.rnex/plugins/<plugin>/skills/`), routed directly through universal `AGENTS.md`.
 
 ---
 
-## Workspace Skills & Workflows
+## Two-Tier Skills Architecture
 
-When a workspace is initialized, Repo Nexus seeds `.rnex/skills/` and `.rnex/workflows/` with specialized agent workflow templates:
+When a workspace is initialized, Repo Nexus maintains core agentic skills in `.rnex/skills/` alongside plugin-owned skills in `.rnex/plugins/<plugin>/skills/`:
 
-| Workflow / Skill | Target Agent Goal |
-|:---|:---|
-| `rnex-cross-repo-feature` | Implement an end-to-end feature across backend, frontend, and shared contracts. |
-| `rnex-workspace-audit` | Audit git status, unpushed branches, dirty files, and schema health. |
-| `rnex-wiki-ingest` | Ingest raw intake documents into atomic Karpathy LLM Wiki pages. |
-| `rnex-wiki-lint` | Validate markdown links, compute backlinks, and update the master index. |
+| Skill | Category | Target Agent Goal | Protocol Path |
+|:---|:---|:---|:---|
+| `clone-sync` | Core | Automated clone, setup, and synchronization of member repositories. | `.rnex/skills/clone-sync.md` |
+| `batch-exec` | Core | Cross-repository batch commands, parallel execution, and verification. | `.rnex/skills/batch-exec.md` |
+| `workspace-status` | Core | Workspace health check, git dirty status, branch tracking, and configuration audit. | `.rnex/skills/workspace-status.md` |
+| `wiki-ingest` | Plugin (`karpathy-llm`) | Ingest raw intake documents into atomic Karpathy LLM Wiki pages. | `.rnex/plugins/karpathy-llm/skills/wiki-ingest.md` |
+| `wiki-lint` | Plugin (`karpathy-llm`) | Validate markdown links, compute backlinks, and update the master index. | `.rnex/plugins/karpathy-llm/skills/wiki-lint.md` |
 
-When provider plugins (such as `copilot`, `claude`, `gemini`, `cursor`, or `windsurf`) are enabled with `prompts: true`, these workflows are automatically mirrored into the respective native prompt pickers (e.g. `.github/prompts/`, `.claude/commands/`, `.gemini/prompts/`, `.cursor/prompts/`, `.windsurf/prompts/`).
+When provider plugins (such as `claude`, `gemini`, `cursor`, or `windsurf`) are enabled with `skills: true`, zero-logic User Skill pointers (`rnex-<skill-name>.md`) are generated directly into the respective native provider directories (`.cursor/rules/`, `.claude/skills/`, `.gemini/skills/`, `.windsurf/rules/`).
 
 ---
 
-## Executing a Cross-Repo Feature
+## Executing a Cross-Repo Batch Feature
 
-### 1. Invoke the Workflow in Your AI Assistant
-In your AI assistant (Copilot Chat, Claude Code, Gemini/Antigravity, Cursor, Windsurf), run or type:
+### 1. Invoke the Skill in Your AI Assistant
+In your AI assistant (Claude Code, Gemini/Antigravity, Cursor, Windsurf), run or type:
 
 ```text
-Follow the workflow in .rnex/workflows/rnex-cross-repo-feature.md to add OAuth2 login support across backend and web-client.
+Follow the skill protocol in .rnex/skills/batch-exec.md to run tests across backend and web-client.
 ```
-*(Or select `rnex-cross-repo-feature` directly from your IDE's prompt picker or slash commands.)*
+*(Or invoke `rnex-batch-exec` directly from your IDE's skill/rules list.)*
 
 ### 2. Autonomous Execution Sequence
 The agent follows a 4-stage protocol:

package/docs/plugins/authoring.md

@@ -41,15 +41,15 @@ A plugin can contain any of the following standard directories:
 
 | Subdirectory | Purpose | Behavior |
 |:---|:---|:---|
-| `workflows/` | Multi-step agent procedural checklists and workflows. | Maintained under `.rnex/plugins/<plugin>/workflows/` and mirrored to provider prompt pickers when `prompts: true`. |
-| `skills/` | Agent skills with `SKILL.md` definitions. | Scoped to `.rnex/plugins/<plugin>/skills/`. Ingested by agents via `AGENTS.md` routing. |
+| `skills/` | Plugin-owned agentic skills (`<skill-name>.md`). | Scoped to `.rnex/plugins/<plugin>/skills/`. Registered as User Skills via `rnex_register_user_skill`. |
+| `rules/` | Coding rules and architectural constraints. | Scoped to `.rnex/plugins/<plugin>/rules/`. |
 | `agents/` | Autonomous subagent declarations (`*.agent.md`). | Scoped to `.rnex/plugins/<plugin>/agents/`. Ingested by agents via `AGENTS.md` subagent routing. |
 | `instructions/` | Markdown guidelines and domain principles. | Scoped to `.rnex/plugins/<plugin>/instructions/`. |
-| `rules/` | Coding rules and architectural constraints. | Scoped to `.rnex/plugins/<plugin>/rules/`. |
+| `hooks/` | Lifecycle hook scripts (`after_sync.sh`, `before_plugin_disable.sh`). | Executed during workspace events. Register/unregister User Skills. |
 | `templates/` | Starter configuration files copied during `rnex sync`. | Copied to mapped paths specified in `templates:`. |
 
 > [!NOTE]
-> All AI assistants share `AGENTS.md` at workspace root. Tools, skills, workflows, and subagents stay inside `.rnex/` and are not mirrored externally. Prompts are exclusively for user invocation through IDE prompt pickers (e.g. `.github/prompts/`, `.claude/commands/`).
+> All AI assistants share `AGENTS.md` at workspace root. Agentic skills, rules, and subagents stay inside `.rnex/` and are not mirrored externally. Active provider plugins expose zero-logic User Skill pointers (`rnex-<name>.md`) in active provider directories (`.cursor/rules/`, `.claude/skills/`, etc.) linking directly to the agentic skill protocols inside `.rnex/`.
 
 ---
 

package/docs/plugins/claude.md

@@ -1,23 +1,23 @@
 ---
 title: Anthropic Claude Plugin
-description: Integrate Anthropic Claude Code with Repo Nexus, mirroring operational prompts into .claude/commands/ for custom slash commands.
+description: Integrate Anthropic Claude Code with Repo Nexus, generating zero-logic user skill pointers in .claude/skills/.
 ---
 
 # Anthropic Claude Plugin
 
 The **`claude`** plugin integrates Anthropic Claude Code CLI (`claude`) and Claude Desktop into a Repo Nexus Virtual Meta-Repo.
 
-All AI assistants in Repo Nexus share [`AGENTS.md`](file:///Users/admin/Projects/ai/repo-nexus/AGENTS.md) at the workspace root as the single universal source of truth for routing. The `claude` plugin mirrors human-facing operational prompts and plugin workflows into `.claude/commands/` and `.claude/prompts/` so they can be triggered directly as custom slash commands (e.g. `/rnex-cross-repo-feature`, `/rnex-workspace-audit`, `/rnex-wiki-ingest`).
+All AI assistants in Repo Nexus share [`AGENTS.md`](file:///Users/admin/Projects/ai/repo-nexus/AGENTS.md) at the workspace root as the single universal source of truth for routing. When enabled, the `claude` plugin generates zero-logic User Skill pointers into `.claude/skills/` (e.g. `rnex-clone-sync.md`, `rnex-batch-exec.md`, `rnex-wiki-ingest.md`).
 
-Operational skills, workflows, rules, and subagents remain cleanly self-contained within `.rnex/` and active plugin directories (`.rnex/plugins/<plugin-name>/`).
+Operational skills, rules, and subagents remain cleanly encapsulated within `.rnex/skills/` and active plugin directories (`.rnex/plugins/<plugin-name>/`).
 
 ---
 
 ## Capabilities
 
-1. **Custom Slash Commands (`.claude/commands/*.md` & `.claude/prompts/*.md`)**:
-   - Mirrors standardized Repo Nexus operational workflows into `.claude/commands/*.md` and `.claude/prompts/*.md`.
-   - In Claude Code CLI, workflows can be invoked directly as custom slash commands (e.g. `/rnex-cross-repo-feature`, `/rnex-workspace-audit`, `/rnex-wiki-ingest`).
+1. **Zero-Logic User Skills (`.claude/skills/rnex-*.md`)**:
+   - Generates lightweight user skill pointers prefixed with `rnex-` in `.claude/skills/`.
+   - Contains zero inline heuristics—redirects Claude directly to the underlying agentic skill protocols inside `.rnex/`.
 
 2. **Native Instruction Support via `CLAUDE.md` & `AGENTS.md`**:
    - Claude Code natively reads `CLAUDE.md` as its primary instructions file; it only falls back to `AGENTS.md` when no `CLAUDE.md` exists in the project.
@@ -25,8 +25,8 @@ Operational skills, workflows, rules, and subagents remain cleanly self-containe
    - If `CLAUDE.md` is absent and `instructions: false`, Claude Code falls back to reading `AGENTS.md` directly.
 
 3. **Subagent & Skill Execution**:
-   - Skills reside under `.rnex/skills/` and `.rnex/plugins/<plugin>/skills/`.
-   - Workflows reside under `.rnex/workflows/` and `.rnex/plugins/<plugin>/workflows/`.
+   - Core skills reside under `.rnex/skills/`.
+   - Plugin skills reside under `.rnex/plugins/<plugin>/skills/`.
    - Subagents reside under `.rnex/agents/` and `.rnex/plugins/<plugin>/agents/` and are invoked as subagents without external mirroring.
 
 ---
@@ -59,16 +59,14 @@ Configure options under `plugins.claude` in `rnex.yaml` or `.local.rnex.yaml`:
 plugins:
   claude:
     instructions: true   # Sync routing instructions to CLAUDE.md (default: true)
-    prompts: true        # Mirror prompts to .claude/commands/ and .claude/prompts/ (default: true)
-    workflows: true      # Mirror active plugin workflows (default: true)
+    skills: true         # Generate user skills in .claude/skills/ (default: true)
     enabled: true        # Set to false to disable
 ```
 
 | Option | Type | Default | Description |
 |:---|:---|:---|:---|
 | `instructions` | boolean | `true` | Merge routing instructions into `CLAUDE.md`. |
-| `prompts` | boolean | `true` | Mirror prompts to `.claude/commands/*.md` and `.claude/prompts/*.md`. |
-| `workflows` | boolean | `true` | Mirror plugin workflows as Claude slash commands. |
+| `skills` | boolean | `true` | Generate user skill pointers in `.claude/skills/rnex-*.md`. |
 | `enabled` | boolean | `true` | Enable or disable the plugin. Overridable in `.local.rnex.yaml`. |
 
 ---
@@ -82,17 +80,16 @@ my-workspace/
 ├── AGENTS.md                          # Universal workspace instructions & routing
 ├── CLAUDE.md                          # Claude-specific instructions (if instructions: true)
 ├── .claude/
-│   ├── commands/                      # Slash command shortcuts
-│   │   ├── rnex-cross-repo-feature.md
-│   │   ├── rnex-workspace-audit.md
-│   │   ├── rnex-wiki-ingest.md
-│   │   └── rnex-wiki-lint.md
-│   └── prompts/                       # Prompts catalog
+│   └── skills/                        # Zero-logic User Skill pointers
+│       ├── rnex-clone-sync.md
+│       ├── rnex-batch-exec.md
+│       ├── rnex-workspace-status.md
+│       ├── rnex-wiki-ingest.md
+│       └── rnex-wiki-lint.md
 ├── .rnex/
-│   ├── skills/                        # Workspace operational skills
-│   ├── workflows/                     # Workspace workflows
+│   ├── skills/                        # Core workspace agentic skills
 │   └── plugins/
-│       └── karpathy-llm/              # Scoped plugin skills, workflows, and agents
+│       └── karpathy-llm/              # Scoped plugin skills and rules
 └── rnex.yaml
 ```
 
@@ -107,6 +104,6 @@ rnex plugin disable claude
 ```
 
 Repo Nexus automatically performs surgical cleanup:
-- Removes auto-generated commands and prompts under `.claude/`.
-- Cleans up any legacy mirrored skills or instructions blocks.
-- Developer-created custom commands, skills, or agents under `.claude/` are preserved.
+- Removes auto-generated user skills under `.claude/skills/rnex-*.md`.
+- Cleans up managed routing instructions in `CLAUDE.md`.
+- Developer-created custom skills under `.claude/` are preserved.

package/docs/plugins/cursor.md

@@ -1,30 +1,31 @@
 ---
 title: Cursor IDE Plugin
-description: Integrate Cursor IDE with Repo Nexus, mirroring operational prompts to .cursor/prompts/.
+description: Integrate Cursor IDE with Repo Nexus, generating zero-logic user skill pointers in .cursor/rules/.
 ---
 
 # Cursor IDE Plugin
 
 The **`cursor`** plugin integrates Cursor IDE and Cursor Agent Mode into a Repo Nexus Virtual Meta-Repo.
 
-All AI assistants in Repo Nexus share [`AGENTS.md`](file:///Users/admin/Projects/ai/repo-nexus/AGENTS.md) at the workspace root as the single universal source of truth for routing. The `cursor` plugin mirrors human-facing operational prompts into `.cursor/prompts/*.md` with standard YAML frontmatter.
+All AI assistants in Repo Nexus share [`AGENTS.md`](file:///Users/admin/Projects/ai/repo-nexus/AGENTS.md) at the workspace root as the single universal source of truth for routing. When enabled, the `cursor` plugin generates zero-logic User Skill pointers into `.cursor/rules/` (e.g. `rnex-clone-sync.md`, `rnex-batch-exec.md`, `rnex-wiki-ingest.md`).
 
-Operational skills, workflows, rules, and subagents remain cleanly self-contained within `.rnex/` and active plugin directories (`.rnex/plugins/<plugin-name>/`).
+Operational skills, rules, and subagents remain cleanly encapsulated within `.rnex/skills/` and active plugin directories (`.rnex/plugins/<plugin-name>/`).
 
 ---
 
 ## Capabilities
 
-1. **Native Instructions & Rules Support**:
-   - Synchronizes Repo Nexus routing instructions to `.cursorrules` and `.cursor/rules/rnex-context.mdc` (`alwaysApply: true`) when `instructions: true` (default).
-   - Preserves user custom rules and custom sections in `.cursorrules`.
+1. **Zero-Logic User Skills (`.cursor/rules/rnex-*.md`)**:
+   - Generates lightweight user skill pointers prefixed with `rnex-` in `.cursor/rules/`.
+   - Contains zero inline heuristics—redirects Cursor directly to the underlying agentic skill protocols inside `.rnex/`.
 
-2. **Prompts Mirroring (`.cursor/prompts/*.md`)**:
-   - Converts standard Repo Nexus operational prompts (`rnex-cross-repo-feature`, `rnex-workspace-audit`) and active plugins' workflows into Cursor prompt templates with YAML frontmatter.
+2. **Native Instructions & Rules Support**:
+   - Synchronizes Repo Nexus routing instructions to `.cursorrules` when `instructions: true` (default).
+   - Preserves user custom rules and custom sections in `.cursorrules`.
 
 3. **Subagent & Skill Execution**:
-   - Skills reside under `.rnex/skills/` and `.rnex/plugins/<plugin>/skills/`.
-   - Workflows reside under `.rnex/workflows/` and `.rnex/plugins/<plugin>/workflows/`.
+   - Core skills reside under `.rnex/skills/`.
+   - Plugin skills reside under `.rnex/plugins/<plugin>/skills/`.
    - Subagents reside under `.rnex/agents/` and `.rnex/plugins/<plugin>/agents/` and are invoked as subagents without external mirroring.
 
 ---
@@ -56,17 +57,15 @@ Configure options under `plugins.cursor` in `rnex.yaml` or `.local.rnex.yaml`:
 ```yaml
 plugins:
   cursor:
-    instructions: true   # Sync routing instructions to .cursorrules & .cursor/rules/ (default: true)
-    prompts: true        # Mirror prompts to .cursor/prompts/ (default: true)
-    workflows: true      # Mirror active plugin workflows as prompts (default: true)
+    instructions: true   # Sync routing instructions to .cursorrules (default: true)
+    skills: true         # Generate user skills in .cursor/rules/ (default: true)
     enabled: true        # Set to false to disable
 ```
 
 | Option | Type | Default | Description |
 |:---|:---|:---|:---|
-| `instructions` | boolean | `true` | Merge routing instructions into `.cursorrules` and `.cursor/rules/`. |
-| `prompts` | boolean | `true` | Mirror prompts to `.cursor/prompts/*.md`. |
-| `workflows` | boolean | `true` | Mirror plugin workflows into `.cursor/prompts/*.md`. |
+| `instructions` | boolean | `true` | Merge routing instructions into `.cursorrules`. |
+| `skills` | boolean | `true` | Generate user skill pointers in `.cursor/rules/rnex-*.md`. |
 | `enabled` | boolean | `true` | Enable or disable the plugin. Overridable in `.local.rnex.yaml`. |
 
 ---
@@ -80,18 +79,16 @@ my-workspace/
 ├── AGENTS.md                          # Universal workspace instructions & routing
 ├── .cursorrules                       # Cursor root instructions (if instructions: true)
 ├── .cursor/
-│   ├── rules/
-│   │   └── rnex-context.mdc           # Cursor alwaysApply routing rule
-│   └── prompts/
-│       ├── rnex-cross-repo-feature.md
-│       ├── rnex-workspace-audit.md
+│   └── rules/                         # Zero-logic User Skill pointers
+│       ├── rnex-clone-sync.md
+│       ├── rnex-batch-exec.md
+│       ├── rnex-workspace-status.md
 │       ├── rnex-wiki-ingest.md
 │       └── rnex-wiki-lint.md
 ├── .rnex/
-│   ├── skills/                        # Workspace operational skills
-│   ├── workflows/                     # Workspace workflows
+│   ├── skills/                        # Core workspace agentic skills
 │   └── plugins/
-│       └── karpathy-llm/              # Scoped plugin skills, workflows, and agents
+│       └── karpathy-llm/              # Scoped plugin skills and rules
 └── rnex.yaml
 ```
 
@@ -106,6 +103,6 @@ rnex plugin disable cursor
 ```
 
 Repo Nexus automatically performs surgical cleanup:
-- Removes auto-generated prompts under `.cursor/prompts/`.
-- Cleans up any legacy mirrored skills, agents, or instructions blocks.
-- Developer-created custom rules and prompts under `.cursor/` are preserved.
+- Removes auto-generated user skills under `.cursor/rules/rnex-*.md`.
+- Cleans up managed routing instructions in `.cursorrules`.
+- Developer-created custom rules under `.cursor/` are preserved.

package/docs/plugins/gemini.md

@@ -1,30 +1,31 @@
 ---
 title: Google Gemini & Antigravity Plugin
-description: Integrate Google Gemini and Antigravity IDE with Repo Nexus, mirroring operational prompts to .gemini/prompts/.
+description: Integrate Google Gemini and Antigravity IDE with Repo Nexus, generating zero-logic user skill pointers in .gemini/skills/.
 ---
 
 # Google Gemini & Antigravity Plugin
 
 The **`gemini`** plugin integrates Google Gemini CLI and Google Antigravity IDE into a Repo Nexus Virtual Meta-Repo.
 
-All AI assistants in Repo Nexus share [`AGENTS.md`](file:///Users/admin/Projects/ai/repo-nexus/AGENTS.md) at the workspace root as the single universal source of truth for routing. The `gemini` plugin mirrors human-facing operational prompts into `.gemini/prompts/*.prompt.md` with standard YAML frontmatter.
+All AI assistants in Repo Nexus share [`AGENTS.md`](file:///Users/admin/Projects/ai/repo-nexus/AGENTS.md) at the workspace root as the single universal source of truth for routing. When enabled, the `gemini` plugin generates zero-logic User Skill pointers into `.gemini/skills/` (e.g. `rnex-clone-sync.md`, `rnex-batch-exec.md`, `rnex-wiki-ingest.md`).
 
-Operational skills, workflows, rules, and subagents remain cleanly self-contained within `.rnex/` and active plugin directories (`.rnex/plugins/<plugin-name>/`).
+Operational skills, rules, and subagents remain cleanly encapsulated within `.rnex/skills/` and active plugin directories (`.rnex/plugins/<plugin-name>/`).
 
 ---
 
 ## Capabilities
 
-1. **Native Instructions Support via `GEMINI.md`**:
+1. **Zero-Logic User Skills (`.gemini/skills/rnex-*.md`)**:
+   - Generates lightweight user skill pointers prefixed with `rnex-` in `.gemini/skills/`.
+   - Contains zero inline heuristics—redirects Gemini/Antigravity directly to the underlying agentic skill protocols inside `.rnex/`.
+
+2. **Native Instructions Support via `GEMINI.md`**:
    - Synchronizes Repo Nexus routing instructions to `GEMINI.md` (when `instructions: true`, default) for native support by the Google Gemini CLI.
    - Preserves user-defined custom sections in `GEMINI.md` across updates.
 
-2. **Prompts Mirroring (`.gemini/prompts/*.prompt.md`)**:
-   - Converts standard Repo Nexus operational prompts (`rnex-cross-repo-feature`, `rnex-workspace-audit`) and active plugins' workflows into Gemini prompt files with YAML frontmatter.
-
 3. **Subagent & Skill Execution**:
-   - Skills reside under `.rnex/skills/` and `.rnex/plugins/<plugin>/skills/`.
-   - Workflows reside under `.rnex/workflows/` and `.rnex/plugins/<plugin>/workflows/`.
+   - Core skills reside under `.rnex/skills/`.
+   - Plugin skills reside under `.rnex/plugins/<plugin>/skills/`.
    - Subagents reside under `.rnex/agents/` and `.rnex/plugins/<plugin>/agents/` and are invoked as subagents without external mirroring.
 
 ---
@@ -57,16 +58,14 @@ Configure options under `plugins.gemini` in `rnex.yaml` or `.local.rnex.yaml`:
 plugins:
   gemini:
     instructions: true   # Sync routing instructions to GEMINI.md (default: true)
-    prompts: true        # Mirror prompts to .gemini/prompts/ (default: true)
-    workflows: true      # Mirror active plugin workflows as prompts (default: true)
+    skills: true         # Generate user skills in .gemini/skills/ (default: true)
     enabled: true        # Set to false to disable
 ```
 
 | Option | Type | Default | Description |
 |:---|:---|:---|:---|
 | `instructions` | boolean | `true` | Merge routing instructions into `GEMINI.md`. |
-| `prompts` | boolean | `true` | Mirror prompts to `.gemini/prompts/*.prompt.md`. |
-| `workflows` | boolean | `true` | Mirror active plugin workflows into `.gemini/prompts/*.prompt.md`. |
+| `skills` | boolean | `true` | Generate user skill pointers in `.gemini/skills/rnex-*.md`. |
 | `enabled` | boolean | `true` | Enable or disable the plugin. Overridable in `.local.rnex.yaml`. |
 
 ---
@@ -80,16 +79,16 @@ my-workspace/
 ├── AGENTS.md                          # Universal workspace instructions & routing
 ├── GEMINI.md                          # Gemini-specific instructions (if instructions: true)
 ├── .gemini/
-│   └── prompts/
-│       ├── rnex-cross-repo-feature.prompt.md
-│       ├── rnex-workspace-audit.prompt.md
-│       ├── rnex-wiki-ingest.prompt.md
-│       └── rnex-wiki-lint.prompt.md
+│   └── skills/                        # Zero-logic User Skill pointers
+│       ├── rnex-clone-sync.md
+│       ├── rnex-batch-exec.md
+│       ├── rnex-workspace-status.md
+│       ├── rnex-wiki-ingest.md
+│       └── rnex-wiki-lint.md
 ├── .rnex/
-│   ├── skills/                        # Workspace operational skills
-│   ├── workflows/                     # Workspace workflows
+│   ├── skills/                        # Core workspace agentic skills
 │   └── plugins/
-│       └── karpathy-llm/              # Scoped plugin skills, workflows, and agents
+│       └── karpathy-llm/              # Scoped plugin skills and rules
 └── rnex.yaml
 ```
 
@@ -104,6 +103,6 @@ rnex plugin disable gemini
 ```
 
 Repo Nexus automatically performs surgical cleanup:
-- Removes auto-generated prompts under `.gemini/prompts/`.
-- Cleans up any legacy mirrored skills, agents, or instructions blocks.
+- Removes auto-generated user skills under `.gemini/skills/rnex-*.md`.
+- Cleans up managed routing instructions in `GEMINI.md`.
 - Developer-created files under `.gemini/` that do not originate from Repo Nexus are preserved.

package/docs/plugins/index.md

@@ -19,12 +19,11 @@ When a plugin is enabled, its rules, instructions, workflows, prompts, and skill
 ### 2. Member Repository Isolation
 Active plugins are mirrored into each active member repository under `repos/<name>/.rnex/plugins/<plugin-name>/`. When an AI assistant operates inside a specific member repository, it has full local access to the workspace's shared instructions and rules without breaking repository autonomy.
 
-### 3. Decoupled Cross-Plugin Asset Discovery
+### 3. Decoupled Two-Tier Skills Architecture
 Repo Nexus decouples functional plugins from AI provider plugins:
-- **Functional Plugins** (such as [`karpathy-llm`](karpathy-llm.md)) declare operational prompts, skills, custom agents, and behavioral guidelines.
-- **Provider Plugins** (such as [`copilot`](copilot.md), [`claude`](claude.md), [`gemini`](gemini.md), [`cursor`](cursor.md), and [`windsurf`](windsurf.md)) automatically scan all active plugins and mirror human-facing prompts and workflows into the provider's native format (`.github/prompts/`, `.claude/commands/`, `.gemini/prompts/`, `.cursor/prompts/`, `.windsurf/prompts/`). Skills, workflows, and subagents remain self-contained under `.rnex/`.
-
-Adding a new workflow or prompt to any functional plugin automatically makes it available across all enabled AI provider tools without manual duplication.
+- **Core Skills (`toolkit/skills/` $\rightarrow$ `.rnex/skills/`)**: Workspace-level capabilities owned by Repo Nexus core.
+- **Plugin Skills (`toolkit/plugins/<plugin>/skills/` $\rightarrow$ `.rnex/plugins/<plugin>/skills/`)**: Scoped domain skills owned exclusively by individual plugins.
+- **User Skills (`rnex-<name>.md`)**: Provider plugins (`claude`, `gemini`, `cursor`, and `windsurf`) generate zero-logic user skill pointers in active provider directories (`.cursor/rules/`, `.claude/skills/`, `.gemini/skills/`, `.windsurf/rules/`) that redirect AI assistants directly to the underlying agentic skills in `.rnex/`.
 
 ### 4. Non-Destructive Delimited Merging
 When plugins update root instruction files (`AGENTS.md`, `CLAUDE.md`, `GEMINI.md`, `.cursorrules`, `.windsurfrules`), Repo Nexus isolates auto-managed routing between explicit delimiters:
@@ -47,11 +46,10 @@ Developer-authored instructions outside of these delimiters are strictly preserv
 
 | Plugin | Category | Description | Documentation |
 |:---|:---|:---|:---|
-| [`claude`](claude.md) | AI Provider | Anthropic Claude Code & Desktop integration: maintains `CLAUDE.md` and mirrors commands and prompts to `.claude/`. | [Read Guide](claude.md) |
-| [`cursor`](cursor.md) | AI Provider | Cursor IDE integration: maintains `.cursorrules`, `.cursor/rules/rnex-context.mdc`, `.cursorignore`, and mirrors prompts to `.cursor/`. | [Read Guide](cursor.md) |
-| [`copilot`](copilot.md) | AI Provider | GitHub Copilot integration: mirrors prompts to `.github/prompts/*.prompt.md` and relies on `AGENTS.md` for routing. | [Read Guide](copilot.md) |
-| [`gemini`](gemini.md) | AI Provider | Google Gemini CLI & Antigravity IDE integration: maintains `GEMINI.md`, `.geminiignore`, and mirrors prompts to `.gemini/`. | [Read Guide](gemini.md) |
-| [`windsurf`](windsurf.md) | AI Provider | Windsurf IDE integration: maintains `.windsurfrules` and mirrors prompts to `.windsurf/`. | [Read Guide](windsurf.md) |
+| [`claude`](claude.md) | AI Provider | Anthropic Claude Code & Desktop integration: maintains `CLAUDE.md` and generates user skills in `.claude/skills/`. | [Read Guide](claude.md) |
+| [`cursor`](cursor.md) | AI Provider | Cursor IDE integration: maintains `.cursorrules`, `.cursorignore`, and generates user skills in `.cursor/rules/`. | [Read Guide](cursor.md) |
+| [`gemini`](gemini.md) | AI Provider | Google Gemini CLI & Antigravity IDE integration: maintains `GEMINI.md`, `.geminiignore`, and generates user skills in `.gemini/skills/`. | [Read Guide](gemini.md) |
+| [`windsurf`](windsurf.md) | AI Provider | Windsurf IDE integration: maintains `.windsurfrules` and generates user skills in `.windsurf/rules/`. | [Read Guide](windsurf.md) |
 | [`karpathy-llm`](karpathy-llm.md) | Knowledge & Rules | Andrej Karpathy's 4 cardinal principles for coding agents, multi-repo context engineering standards, and autonomous zero-dependency LLM Wiki architecture (`./docs`, `.rnex/wiki`). | [Read Guide](karpathy-llm.md) |
 
 ---