Problem / Use Case

Enabling developers advanced workflows like https://github.com/ChristopherKahler/paul and https://github.com/open-gsd/gsd-core with custom slash commands, deep template systems, and structured state management.

Current limitation: Kimi Code does not recognize or load custom slash commands from ~/.kimi/commands/ or any user-defined directory. This means:

• Developers cannot invoke /{custom-command} natively in Kimi Code
• The entire structured workflow must be manually copy-pasted or referenced via complex prompts, defeating the purpose of a repeatable, state-managed system
• Context rot is inevitable without loop-enforced commands that read/write state files automatically

What Claude Code does (and why it matters)

Claude Code auto-loads markdown files from ~/.claude/commands/ as first-class slash commands. Each .md file becomes an executable command with its content injected as a prompt template. This enables:

• Framework distribution via npx install or git clone
• In-session state management commands read/write project state without user intervention
• Quality enforcement structured workflows are embedded in the command itself, not optional

Proposed Solution

Add a custom slash command directory to Kimi Code, similar to:

~/.kimi/commands/          # global custom commands

Command file format (backward-compatible):

---
name: {custom-command}
description: Create an executable {custom-command} with acceptance criteria
---

Why this matters for Kimi Code adoption

1. Enterprise/team workflows Teams can distribute standardised command sets for compliance, security review, or architecture governance
2. Framework ecosystem Enables PAUL, GSD, and other structured AI development frameworks to be adopted natively on Kimi
3. Differentiation from raw chat Custom commands elevate Kimi Code from "smart autocomplete" to "repeatable workflow engine"
4. Stateful sessions Commands that read/write project state require native integration, not prompt hacks

Request

Please implement custom slash command loading in Kimi Code, prioritizing:

1. Directory scanning (~/.kimi/commands/, ./.kimi/commands/)
2. Markdown file format with YAML frontmatter (name, description)
3. Integration with Kimi's slash command palette (/)
4. Project-local file system access from within command execution

This would make Kimi Code the preferred platform for structured, stateful AI-assisted development.