Files
leo-claude-mktplace/plugins/code-sentinel/commands/sentinel.md
lmiranda 5bbcee06ac fix(commands): wire dispatch files to real sub-command names via Skill tool
All dispatch files referenced display names (e.g., `/doc audit`) that don't
correspond to real slash commands. Claude Code resolves commands by filename,
so `/doc audit` loads `doc.md` with $ARGUMENTS but never routes to `doc-audit.md`.

Changes:
- Updated 24 dispatch files with explicit Skill tool routing
- Added "Command to Invoke" column with plugin-prefixed names
- Added $ARGUMENTS matching for automatic routing
- RFC dispatch file uses inline handling (no separate command files)
- Updated COMMANDS-CHEATSHEET.md with invocation methods
- Updated MIGRATION-v9.md with command name mapping explanation

Affects: 25 dispatch files across 12 plugins (core, data, saas, ops, debug)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
2026-02-08 21:04:24 -05:00

33 lines
1.3 KiB
Markdown

---
name: sentinel
description: Security scanning and code refactoring — type /sentinel <action> for commands
---
# /sentinel
Security scanning and safe code refactoring tools.
When invoked without a sub-command or with `$ARGUMENTS`, handle as follows:
## Available Commands
| Action | Command to Invoke | Description |
|--------|-------------------|-------------|
| `scan` | `/code-sentinel:sentinel-scan` | Full security audit (SQL injection, XSS, secrets, etc.) |
| `refactor` | `/code-sentinel:sentinel-refactor` | Apply refactoring patterns to improve code |
| `refactor-dry` | `/code-sentinel:sentinel-refactor-dry` | Preview refactoring without applying changes |
## Routing
If `$ARGUMENTS` is provided (e.g., user typed `/sentinel scan`):
1. Match the first word of `$ARGUMENTS` against the **Action** column above
2. **Invoke the corresponding command** from the "Command to Invoke" column using the Skill tool
3. Pass any remaining arguments to the invoked command
If no arguments provided:
1. Display the Available Commands table
2. Ask: "Which action would you like to run?"
3. When the user responds, invoke the matching command using the Skill tool
**Note:** Commands can also be invoked directly using their plugin-prefixed names (e.g., `/code-sentinel:sentinel-scan`)