dsh-prompt-only-forge
Run the following command in DeepSeek Harness:
dsh plugin install shengyvself/dsh-prompt-only-forge
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install shengyvself/dsh-prompt-only-forge in the DeepSeek Harness terminal to install; the source is at https://github.com/shengyvself/dsh-prompt-only-forge. Restart dsh web after installation.
About this plugin
Polishing a prompt should not require spinning up a sub-agent, making a network call, and running a single-shot LLM rewrite. dsh-prompt-only-forge strips the task to its minimum: click the ✨ slot in the composer, a polish-prompt template wraps your current draft, and the main agent of your session—which naturally holds all history, tools, and file access—produces the final text. The plugin itself makes no network requests, no model calls, and no disk writes; it never sends on your behalf, and you decide whether to press Enter.
The design follows three rules: inject only, never send, zero dependencies. Your draft is preserved verbatim and slotted into the template placeholder; clicking ✨ when the template is already present simply shows a gentle toast instead of re-wrapping; if inputActions is unavailable, an explicit red toast appears rather than a silent fallback. The plugin registers on DSH's official conversation.input.right slot, with no wild DOM injection and no dependency on dsh-better-sidebar.
It is built for developers and Agent orchestrators who write or iterate prompts frequently in DSH, and for anyone who wants the polish step to be side-effect-free, predictable, and auditable. It does not interrupt your workflow or introduce a new failure surface—just click ✨, press Enter, and let the main agent do the rest.
Screenshots
Use Cases
- Click ✨ after drafting a prompt to polish without composing a manual instruction
- Let the main agent rewrite a prompt with full session context
- Keep the polish step offline, model-free, and non-sending
- Preserve an existing draft verbatim when injecting the template
Best For
- Developers who write or iterate prompts frequently in DSH
- Agent orchestrators who want a predictable, auditable, low-failure polish pipeline
- Prompt engineers who value zero-dependency, minimal side-effect workflows
Related Plugins
A unified suite combining hot runtime injection, task-aware thinking-mode routing, and a graded session protocol with red-team gates to sustain model diligence across long-horizon inference.
ModLens is a vision plugin for DeepSeek Harness that gives text-only models sight by reading images pasted directly into chat, with zero-config setup and multiple vision engines.
On-demand vision for text-only DeepSeek Harness agents: built-in free keyless vision chain and 14 vision tools, routing image turns as tool calls to vision models with pixel fidelity, no Python needed, one-command install.
Give text-only models in DeepSeek Harness eyes, enabling image Q&A, long-screenshot OCR, UI restoration, and GUI visual tasks.