Slash Commands
Fletch doesn’t have a separate command palette — instead, the composer is where you
reach for actions as you write a message, using /, @, and # triggers. All three
work in an agent’s chat and in the new-agent draft composer.
Slash commands
Section titled “Slash commands”Type / at the start of the composer to open the command autocomplete. What’s listed
depends on the agent; today Claude Code has the full set, and other agents expose
none. The list has four sources:
Built-in commands. /help, /compact, and /init are forwarded to the agent
verbatim. A second group of Claude built-ins can’t be forwarded, so Fletch handles them
itself: /doctor and /mcp run the real claude subcommand and show its output;
/cost shows this session’s token usage; /config opens Settings; /clear starts a
fresh session; /resume opens the session history. Command output renders as a readable
block in the transcript.
Your custom commands. Fletch discovers the same command files Claude Code does —
*.md files under ~/.claude/commands and <project>/.claude/commands — and offers
them in the autocomplete.
Claude skills and plugin commands. Skills from ~/.claude/skills and
<project>/.claude/skills appear under their names, and commands from installed Claude
plugins appear namespaced as /<plugin>:<command>. On a name clash, the more intentional
entry wins: project command over project skill, over user command, over user skill, over
plugin command — and nothing can shadow a built-in.
Your Fletch skills. Skills from your Settings › Skills
library are invocable as /<skill-name>. Picking one
attaches the skill to the agent and rewrites your message into a follow-it-now prompt;
any text after the command is passed along as arguments.
Attaching files with @
Section titled “Attaching files with @”Type @ to mention and attach files from the workspace. You can also drag-and-drop or
browse for attachments. In a workflow launch, attachments are delivered to the run’s
first step.
Referencing PRs with #
Section titled “Referencing PRs with #”Type # to mention a pull request, pulling its context into your message.
Composer chips
Section titled “Composer chips”Alongside the text field, the composer has a thinking-effort chip (the levels offered depend on the selected model — some models go beyond high), an auto-edit toggle, and a context-window usage meter — each can be shown or hidden under Settings › General.