dsh-anchored-monitor
Run the following command in DeepSeek Harness:
dsh plugin install Aik358/dsh-anchored-monitor
Paste the following prompt into your AI chat to install this plugin:
To install this plugin in DeepSeek Harness, please visit https://github.com/Aik358/dsh-anchored-monitor to obtain the source code and follow the documentation for deployment.
About this plugin
When building complex Agent workflows with DeepSeek V4 Pro, the model often drifts from a highly focused "We" mode into a scattered "let me" mode due to context pollution, a state that rarely self-corrects. dsh-anchored-monitor acts as an invisible whip, continuously analyzing reasoning fingerprints to ensure the model remains in a high-performance state.
At its core, the plugin features a liquid-glass UI panel (including a fun "Liang-o-meter" meme skin) that visualizes thinking intensity with real-time, ECG-style curves. Upon detecting trajectory drift, it triggers a tiered intervention system: from gentle suggestive hints (L1) to auto-canceling and resetting to a minimal persona (L2), up to recommending a restart (L3), seamlessly auto-continuing tasks without manual takeover.
This tool is tailor-made for developers and Agent builders deeply invested in DeepSeek Harness. If you are frustrated by your model losing focus during long conversations, this zero-setup monitor acts as a tireless supervisor, keeping your Agent operating at peak reasoning and execution capability.
Screenshots
Use Cases
- Monitor the decay and drift of model reasoning capabilities in long conversations.
- Automatically correct the model when it deviates from the focused persona mode.
- Visualize the thinking intensity and chain-of-thought health of LLMs in real time.
Best For
- Agent developers deeply utilizing DeepSeek V4 Pro.
- AI engineers requiring stable outputs in complex workflows.
- Prompt engineers dedicated to optimizing LLM reasoning performance and focus.
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.