dsh-approve-for-me
Run the following command in DeepSeek Harness:
dsh plugin install shifan3/dsh-approve-for-me
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install shifan3/dsh-approve-for-me in the DSH terminal to install the plugin; the full source code is available at https://github.com/shifan3/dsh-approve-for-me
About this plugin
Every tool call in a DeepSeek Harness session can trigger an approval prompt. Clicking through repetitive low-risk confirmations breaks flow, while genuinely destructive commands easily get buried in the noise. dsh-approve-for-me automates the former and safeguards the latter.
The plugin prepends an LLM safety reviewer to the approval chain. It feeds the most recent ~12 k characters of conversation context together with the requested action to the model, which responds with a single word: ALLOW or REJECT. If the action matches a high-risk pattern such as rm -rf, drop table, git push --force, or a fork bomb, the request is immediately handed back to the interactive user prompt and the model is never consulted. On timeout, unparseable output, or a missing model service, the plugin falls through to the human prompt as well—never silently allowing, never silently rejecting.
Ideal for DSH users who want frictionless approval of routine tool calls while keeping a mandatory human-in-the-loop gate for any destructive or irreversible operation. Activation is a single enabled flag; no changes to the approval policy enum are required, and standard audit events continue to be recorded.
Use Cases
- Auto-approve routine read-only tool calls to eliminate repetitive confirmation prompts
- Route destructive commands such as rm -rf or drop table back to a human approver automatically
- Enforce a consistent approval policy across shared DSH environments to reduce accidental damage
Best For
- Developers frequently interrupted by approval prompts during high-throughput DSH sessions
- Operations engineers who want human-in-the-loop for critical actions while reducing routine confirmations
- DSH team admins who need a uniform, auditable approval policy across shared environments
Related Plugins
A network security red-team benchmark plugin for DeepSeek-V4.1/Flash featuring dual-layer kernel injection and zero-tool architecture for controlled model safety assessment.
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.