← DeepSeek Harness Plugins
DeepSeek Harness PluginManifest Valid

dsh-cot-anchor

Extracts established conclusions from recent reasoning to inject after tool outputs and cuts off repetitive or degraded CoT loops.

Developer Tools
0GitHub Stars0ForksUpdated2026-10-14

Install

$ dsh plugin --profile <你的profile> add github:JiewiW/dsh-cot-anchor#v0.2.1

Plugin Overview & Capabilities

AI-assisted organization based on the public repository snapshot. The content must be grounded in source evidence and does not replace compatibility or security verification.

source-grounded
dsh-cot-anchor is a DeepSeek Harness plugin designed to combat model reasoning stagnation. It automatically extracts key established conclusions from the model's most recent reasoning trace and injects them back into the context after tool execution to prevent re-deriving already proven facts. In addition, it provides generation-time soft-cutting to detect and interrupt pseudo tool calls, verbatim repetitions, churn, and runaway number streams.

Key Capabilities

  • Automatic conclusion anchor extraction and injection following tool boundary results without invalidating prompt cache
  • Real-time soft-cutting generation checks to abort churn, exact string loops, pseudo-tool formatting, and runaway digit sequences
  • Configurable anchor constraints including character limits, point caps, and minimum reasoning length
  • Optional execution discipline reminders injected via runtime context snapshots
  • Optional LLM-based reasoning summarization and local CoT sample harvesting

Useful For

  • Preventing the model from endlessly re-analyzing already solved facts during complex multi-step tasks
  • Catching degraded reasoning loops such as verbatim phrase repeats or pseudo tool call syntax before tokens are wasted
  • Optimizing agent prompt cache efficiency while preserving intermediate thought milestones

Who It Fits

  • DeepSeek Harness users running tool-heavy reasoning agents
  • Developers seeking to reduce token waste and reasoning churn during long-context execution

Documented Limitations

  • Soft-cut generation interruption requires specific DSH kernel support (0.1.5-rc.1/rc.2 or backported hook)
  • Only handles single-stream reasoning degradation and cannot solve cross-step wandering across multiple turns
  • Heuristic transition sentence interruption may trigger false positives if thresholds are set too aggressively

DSH Compatibility

Version-specific runtime evidence collected by DSH Plugin. A missing result means we have not tested that combination yet.

Not tested yetNo runtime compatibility tests have been published yet.

Security Signals

Objective signals discovered from package metadata and source inspection. These are not a guarantee that a plugin is safe.

Dependency Counts

package.json declares 0 runtime, 0 development, 1 peer, and 0 optional dependencies.

info
Dsh Bundle Declared

package.json declares DSH bundle metadata.

info
License Declared

GitHub reports the repository license as MIT.

info
Package Manifest Available

A root package.json was captured and can be inspected by the registry.

info
Source Available

Public GitHub source metadata is available for this registry snapshot.

info

Source & Registry Notes

Public provenance, Registry classification, and the latest source check for this entry, kept separate from runtime verification.

Source repository
JiewiW/dsh-cot-anchor
Registry source
GitHub · dsh-plugin topic
Registry classification
Plugin
Source checked
1b01d8e · 2026-10-14

This project is independently indexed from public source information. DSH Plugin is not affiliated with DeepSeek or the plugin author. Always check the author repository before installation.

Repository Activity

GitHub stars
0GitHub stars
Forks
0
Open issues
0
Last commit
2026-10-14
Last release
2026-10-14
For maintainers

Maintaining this plugin?

This listing is generated from public repository data. If you maintain this project, you can review the information and share this listing with your users if you find it useful.

Add to README
Listed on DSHPlugin.app