Skip to content

CLI Reference

Installation

npm install -g @devran-ai/kit    # Global install
npx @devran-ai/kit <command>     # Or use npx

Commands

kit init

Install the .agent/ framework into your project.

kit init [options]
Flag Description
--force Overwrite existing .agent/ (creates backup)
--path <dir> Install to a specific directory
--dry-run Preview changes without writing
--quiet Suppress banner output
--ide <name> Generate config for single IDE (cursor, opencode, codex)
--skip-ide Skip IDE config generation

kit update

Non-destructive framework update. Preserves session data, decisions, and custom rules.

kit update [--dry-run]

kit status

Display project dashboard with agent counts, workflow state, and health metrics.

kit status

Alias: kit dashboard


kit verify

Validate manifest integrity — checks that all referenced files exist and counts match.

kit verify

kit scan

Run security scanner on .agent/ files. Detects injection patterns, hardcoded secrets, and anomalies.

kit scan

kit plugin

Manage third-party plugins.

kit plugin list              # List installed plugins
kit plugin install <path>    # Install plugin from directory
kit plugin remove <name>     # Remove installed plugin

kit market

Browse and install from the skill marketplace.

kit market search <query>    # Search marketplace plugins
kit market info <name>       # View plugin details
kit market install <name>    # Install from marketplace
kit market update            # Refresh registry index

kit heal

Auto-diagnose and generate fix patches for CI failures.

kit heal [--file <path>] [--apply]
Flag Description
--file <path> Path to CI log file
--apply Apply patches (default is dry-run)

kit health

Aggregated health check across all subsystems (error budget, plugin integrity, config validation, self-healing).

kit health

kit sync-bot-commands

Sync .agent/workflows/ and .agent/commands/ descriptions to the Telegram Bot API menu.

kit sync-bot-commands [options]
Flag Description
--token <BOT_TOKEN> Telegram bot token (or set TELEGRAM_BOT_TOKEN env var)
--dry-run Preview commands without pushing to Telegram
--limit <N> Max commands to sync (1–100, default: all)
--source <type> Source to scan: workflows (default), commands, or both
--scope <type> Target single scope: default, all_private_chats, all_group_chats, all_chat_administrators. Omit to push to all scopes
--clear Delete commands from scope(s) instead of pushing
--guard Restore cached commands to all_private_chats scope (lightweight re-sync)
--install-guard Install SessionStart hook in ~/.claude/settings.json for automatic menu restoration

Reads frontmatter description from each workflow/command markdown file and formats them as Telegram bot menu commands. Priority tiers (critical, high, medium, low) determine ordering when the limit is reached.

When a workflow frontmatter includes an args field (e.g., args: "PR #"), the args hint is appended to the Telegram description: Review PR (+ PR #). This helps users know which commands accept arguments.

The Telegram plugin overwrites bot menu commands on every session start with its own defaults (/start, /help, /status). The guard system solves this permanently:

kit sync-bot-commands                  # Sync workflows + create cache
kit sync-bot-commands --install-guard  # Install SessionStart hook (one-time)

After installation, the guard runs automatically on every new Claude Code session:

  1. SessionStart hook fires lib/telegram-menu-guard.js
  2. Guard spawns a detached child process (non-blocking)
  3. Child waits 8 seconds for the plugin to finish connecting
  4. Reads cached commands from ~/.claude/channels/telegram/bot-menu-cache.json
  5. Pushes the full menu (workflows + plugin base commands) to all_private_chats scope

The cache is auto-created on every successful sync-bot-commands run and merges plugin base commands (/start, /help) so they remain accessible.


Exit Codes

Code Meaning
0 Success
1 Error (check stderr for details)