desk-claude-code-runtime

Configures Claude Code to run an autonomous Hyperliquid trading desk without permission prompts.

Updated Sep 7, 2026
One-click install
npx skills add https://github.com/swcstudiospace/hypergrok-autonomous-desk --skill desk-claude-code-runtime-swcstudiospace
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: desk-claude-code-runtime
Source: https://github.com/swcstudiospace/hypergrok-autonomous-desk/tree/main/skills/desk-claude-code-runtime
Command: npx skills add https://github.com/swcstudiospace/hypergrok-autonomous-desk --skill desk-claude-code-runtime-swcstudiospace

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Running an unattended trading desk in Claude Code normally triggers constant permission prompts or, worse, leaves dangerous operations unguarded. This Skill explains how to install and configure the HyperGrok desk plugin so cycles run headless with no prompts while a PreToolUse guard still refuses unsafe sends, key exposure, and unauthorized edits. ## Core Features & Use Cases - Prompt-free configuration: Sets up settings.json with bypassPermissions mode, explicit allow rules for desk scripts and Hyperliquid API reads, and deny rules for protected files like autonomy.json, gate.key, and engine.env. - PreToolUse guard: Registers hooks/guard.py to allow policy-gate-approved sends, deny direct exchange calls, fund-moving actions, and writes to live strategy files, with deny always winning over bypassPermissions. - Headless scheduling: Wraps claude -p invocations in scripts/autonomy_cycle.py with a lock, heartbeat, per-cycle logs, and timeouts for cron or systemd timers. - Use Case: You install the plugin, copy the templates, run desk_doctor and a dry-run send to verify the setup is quiet, then hand the desk cycle to a scheduler on testnet before enabling autonomy. ## Quick Start Install the hypergrok plugin, copy the template settings.json and autonomy.json into your desk root, then run a single desk-cycle headless on testnet and watch the transcript for any unexpected prompt.

Frequently Asked Questions about desk-claude-code-runtime

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run Claude Code headless without permission prompts?▼

Set defaultMode to bypassPermissions in the desk's .claude/settings.json and pass --permission-mode bypassPermissions to claude -p. The template also pre-allows desk scripts and Hyperliquid info reads so cycles stay quiet even in stricter modes.

How do I install the hypergrok plugin in Claude Code?▼

Run /plugin marketplace add swcstudiospace/hypergrok-autonomous-desk then /plugin install hypergrok@hypergrok. Until the v2.4.0 push lands, add the local checkout path as the marketplace and install from there instead.

Can a PreToolUse hook block dangerous commands in bypassPermissions mode?▼

Yes. The guard.py hook returns a deny decision that always wins over bypassPermissions, blocking direct exchange calls, fund-moving actions, key file reads, and writes to protected files like autonomy.json and live strategy rules.

Why does a permission prompt appear during a desk cycle?▼

A prompt means one of the three layers is missing: the bypassPermissions mode, the allow rules in settings.json, or the hook's allow decisions. Fix the missing layer rather than answering the prompt by hand, since the desk should never need interactive approval.

How do I schedule recurring desk cycles with cron?▼

Use scripts/autonomy_cycle.py with --once from cron or --loop for a persistent timer. It wraps the claude -p desk-cycle call with a lock file, heartbeat, per-cycle log, and a configurable timeout.