dsh-inject-system-prompt
Run the following command in DeepSeek Harness:
dsh plugin install lovstudio/dsh-inject-system-prompt
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install lovstudio/dsh-inject-system-prompt in your DeepSeek Harness workspace; source repository: https://github.com/lovstudio/dsh-inject-system-prompt
About this plugin
When building agents on DeepSeek Harness, you often want to slip an extra instruction into the system prompt—adjusting the persona, adding tool-usage guidance—without touching the core codebase. dsh-inject-system-prompt closes that gap: it registers a single system-prompt section through the harness registry so you can inject, reorder, or even replace the entire prompt purely from configuration.
Under the hood the plugin makes one call—ctx.systemPrompt.section()—to register an entry. Every knob, including section text, display name, sort order, and the complete flag that promotes this section to the sole prompt, lives as a plain key-value pair in cordis.patch.yml. Ordering follows a banding scheme (0 for persona, 100–199 for tool guidance) so you can slot your text exactly where it belongs. Leave the text empty and the section vanishes; the bundle stays completely inert until you configure it.
It is aimed at developers already running DeepSeek Harness who want the least-invasive way to customize an agent's prompt. There are no runtime dependencies and no build step: change one config line to activate, blank it out to revert—no forking, no recompilation, no version pinning.
Use Cases
- Inject a persona tweak or role constraint into an existing agent prompt
- Append tool-usage guidelines or behavioral rules to a harness agent
- Temporarily override the complete system prompt for A/B testing
Best For
- Developers building agents on DeepSeek Harness
- Engineers who need a non-invasive way to customize agent prompts
- DevOps teams managing multiple agent profiles
Related Plugins
A method pack that makes AI coding agents plan against your real baseline, prove completion with fresh evidence, and reduce reworks and unsafe changes.
Turns the DeepSeek Harness session into a captain that builds a durable sub-agent team, splits goals into dependency-aware tasks, and coordinates work via direct messages and a live Web UI.
Gives coding agents design judgment, letting Claude Code, Cursor, and 70+ agents generate and iterate high-quality UI, presentations, and graphics right from the terminal.
Run the Pi ecosystem's plugins on DeepSeek Harness, unmodified, via a compatibility layer that implements Pi's public extension ABI on DSH's native services.