Slash Commands

Inside the TUI, type / to open the command palette. Me Write Code ships built-in commands and reads user-authored markdown commands from ~/.mewrite/agent/commands/ and project-local .mewrite/commands/.

Built-in commands

Command	What it does
/login /logout	OAuth flow for any provider
/model	Switch model mid-session, transcript reformatted in place

| /settings | Edit ~/.mewrite/agent/settings.json in your $EDITOR | | /new /resume /tree /fork | Session lifecycle and branching | | /compact | Manual context compaction (auto on overflow) | | /copy /export /share | Copy transcript / export JSON / share link | | /plan /act | Switch between plan and edit modes | | /tokens | Live token / cost breakdown | | /repomap | Show the current repo map (PageRank-ranked symbols) | | /architect | Architect/editor split mode | | /checkpoint /rollback N | Shadow-git snapshots | | /memory | cavemem queries: search, save, consolidate, sync | | /help /hotkeys /changelog | Self-explanatory |

Default user commands shipped with Me Write Code

These ship in ~/.mewrite/agent/commands/ after first run. You can edit or delete any of them.

Command	What it does
/commit	Stage changes, propose a conventional-commit message, commit
/test	Run the test suite, summarize failures
/review	Read the diff vs main and produce a code-review summary
/explain	Explain the file/symbol under cursor
/fix-types	Walk type errors and propose fixes
/perf	Profile the cwd app and suggest optimizations
/sec-review	Run a security audit pass on the diff
/clean	Delete dead imports, unused vars
/log	Append a session note to CHANGELOG.md
/migrate	Run a recipe from .mewrite/recipes/

Authoring a slash command

Create ~/.mewrite/agent/commands/my-cmd.md:

---
description: "Run prettier on the staged files and re-stage"
argument-hint: "[--all]"
allowed-tools: [Bash, Edit]
model: claude-haiku-4
---

You are a code-formatter. Run prettier on every file in `$ARGUMENTS` (default
to staged files: `!`git diff --cached --name-only``). Re-stage the changes.

Frontmatter keys (full list — superset of Claude Code):

Key	Purpose
description	Auto-loaded into the command palette
argument-hint	Shown next to the command name
arguments	Typed arg schema (validated before run)
allowed-tools	Tool allowlist for this command
disallowed-tools	Tool denylist
disable-model-invocation	Hide from auto-suggestion
user-invocable	Default true
model	Override session model for this command
effort	low/medium/high thinking
context: fork	Fork into a sub-session
agent	Dispatch to a named subagent
hooks	Inline hook overrides
paths	Limit the command to certain glob paths
shell	Inline shell preprocessing language

Substitutions

• $ARGUMENTS — full argv after the command
• $0, $1, ... — positional args
• ${CAVE_SESSION_ID}, ${CAVE_SKILL_DIR}, ${CAVE_EFFORT} — session vars
• !`cmd` — runs cmd at command-load time, substitutes stdout

Hot reload

Me Write Code watches ~/.mewrite/agent/commands/ and .mewrite/commands/ for changes. Save the file, the new version is live on the next /.

Project vs user vs plugin scope

Scope	Path	When to use
Project	.mewrite/commands/*.md	Lives in repo; team shares it
User	~/.mewrite/agent/commands/*.md	Personal preferences
Plugin	node_modules/<plugin>/commands/*.md	Installed via mewrite plugin install

User overrides project overrides plugin. Conflicts surface in mewrite doctor.

Importing from Claude Code

cp ~/.claude/commands/*.md ~/.mewrite/agent/commands/

The frontmatter formats are identical. Paste-and-go.