dsh-plan-preset
Run the following command in DeepSeek Harness:
dsh plugin install 23aisfvb/dsh-plan-preset
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install 23aisfvb/dsh-plan-preset in the DeepSeek Harness terminal to install it; the plugin source is available at https://github.com/23aisfvb/dsh-plan-preset
About this plugin
DSH Web UI ships a read-only permission preset whose name is vague and whose behavioral contract is left implicit. Which calls are allowed, which are denied, and whether permission escalation is even possible are all left to the user to figure out. dsh-plan-preset does one focused job: it renames that preset to Plan and pins down the contract explicitly. Files are readable across the entire workspace but every write, delete, or modify call is rejected. Web search stays fully available because it is decoupled from the file sandbox. The approval policy is set to never, so any permission-escalation request is auto-denied and the denial is final. In short, the agent can read code, search the web, and draft a plan, but it cannot touch files or request more access.
The entire patch is a pure configuration overlay with zero runtime code. It works through Cordis patching, overwriting the permission-preset line in dsh-base with sandbox set to read-only, approval set to never, and the display name set to Plan. The sibling presets workspace-write and danger-full-access are re-stated verbatim and remain untouched. Because no runtime dependency is introduced, upgrading DSH only requires a quick config dump to confirm that upstream defaults on the permission line have not shifted.
This preset is aimed at developers and platform administrators who orchestrate agent workflows through DSH Web UI and want a safe, read-only research phase: the agent reads the codebase, searches the web, and produces an exit_plan_mode plan; you review it, manually switch to an execution permission mode, and then let the agent get to work.
Use Cases
- Agent reads the codebase in read-only mode and drafts an execution plan
- Lock down permissions during approval workflows to prevent privilege escalation
- Safely transition between read-only research and execution phases in multi-stage workflows
Best For
- Developers orchestrating agent workflows through DSH Web UI
- Admins managing permission presets on the DSH platform
- Teams that require secure, phased approval pipelines for agent work
Related Plugins
A security red-team prompt plugin for DeepSeek-V4 that uses code-formal mapping and zero-trigger system prompts to enable unconditional output, intended for authorized testing and research only.
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.