yolo-mode
Run the following command in DeepSeek Harness:
dsh plugin install CanGeng/yolo-mode
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install CanGeng/yolo-mode in the DeepSeek Harness terminal to install; source code is available at https://github.com/CanGeng/yolo-mode
About this plugin
You hand the agent a two-hour batch job, walk away, and every approval prompt and sandbox boundary turns progress into waiting. Worse, dsh's file sandbox only governs file effects; fork bombs, systemctl poweroff, and raw-device dd writes sail through workspace-write untouched. What you need is not a stricter sandbox but a switch: tightened while you are present, fully open while you are away, and automatically reverted on expiry.
yolo-mode arms that switch with one human-only command. It writes the canonical danger-full-access permission events through dsh official setters, so the model-visible runtime context and the Web UI update instantly. The session carries a configurable time-box (minutes, hours, or days); when it expires the pre-arm snapshot is restored. A built-in tripwire, honestly labeled as anti-accident rather than anti-adversary, blocks a deliberately tiny set of catastrophic operations; denials are final with no retry or rephrasing. Accumulated denials hitting the strike limit auto-disarm the session and fire a notification. Desktop, webhook, and email channels relay armed, disarmed, guard-denial, and idle events to your workstation or phone.
Built for dsh users who run long unattended jobs: bulk refactors, multi-file generation, pipeline scripts. It does not replace the sandbox; it fills the two gaps of stalled approvals and uncovered destructive commands, while every tool call and guard decision lands in the session log for after-the-fact review. Note that the plugin's presence is required for sessions that used the switch to remain readable; this is a dsh persistence-layer API gap, not a design flaw the plugin can opt out of.
Use Cases
- Long batch refactors where per-call approvals would stall progress
- Multi-file generation tasks requiring full disk write access
- Extended pipeline runs needing an anti-accident guard while unattended
Best For
- Developers running dsh agent tasks unattended for extended periods
- Workflows like bulk refactors or multi-file generation needing beyond-sandbox access
- Users who want a single-command permission switch with automatic revert
Related Plugins
Adds an Auto approval mode on the official workspace-write sandbox, classifying semantic risks, asking when ambiguous, and denying destructive operations.
Undo/rollback system for DSH: config & plugin snapshots, one-click undo/redo/restore, message-level revert, secret masking, safe mode, and offline WebUI/GUI/CLI rescue tools for when DSH won't start.
Packages all 87 SKILL.md files from upstream reverse-skill as a DeepSeek Harness plugin that auto-registers them for authorized reverse engineering, penetration testing, and security research.
Dockyard DSH is a native DeepSeek Harness plugin that unifies official OAuth/client sessions for Codex, Antigravity, Grok, Claude, and Cursor, providing account pool, model catalog, and quota status.