deepseekplugins

DeepSeek Harness plugin

dsh-v4-anchor

面向 DeepSeek Harness Standard 的 DeepSeek V4 首轮 RL 锚定插件，promotion 后恢复完整能力与 Skill。

Tools & Capabilities
Jump to install

Source facts

Repository
Starfie1d1272/dsh-v4-anchor
Latest update
Aug 18, 2026
Category
Tools & Capabilities
GitHub stars
2
Format
plugin
Catalog evidence
Upstream dsh.bundle evidence
Evidence path
package.json#dsh.bundle
Checked against
0.1.0-rc.8
Upstream check date
2026-08-20

This evidence comes from the upstream catalog. This site has not installed, run, or security-reviewed the plugin.

Install

Start with a prompt that asks an agent to review the GitHub repository and source. Switch to the command if you want to install it yourself.

Copy this prompt into DSH, Codex, or another agent and ask it to review the GitHub repository and source first.

Do not install or run any commands yet. Read this plugin's GitHub repository, README, and relevant source code. Then answer the questions below clearly and directly so I can decide whether it fits my needs:

1. What is this plugin, and what problem does it solve?
2. Who is it for, and what are its typical use cases?
3. How is it used after installation? Include one minimal example.
4. What known limitations or privacy, security, compatibility, or maintenance risks does it have?
5. Give a clear recommendation: recommend, conditionally recommend, or do not recommend, with reasons.

Distinguish statements documented by the repository, inferences from source code, and unknowns. If evidence is insufficient, say so explicitly. Do not guess or simply repeat the README.

GitHub: https://github.com/Starfie1d1272/dsh-v4-anchor
Plugin: dsh-v4-anchor
Author: Starfie1d1272

Check the source files

Read the README and other files from this plugin directory before installing.

File explorer4 files
README.en.mdSource · read only
README language

dsh-v4-anchor

中文

A deliberately small DeepSeek Harness plugin that preserves one narrow, evidence-backed intervention:

> Use an RL-shaped bootstrap for the first request of an eligible DeepSeek V4 Standard session, then restore the full Standard capability surface after the first durable tool call.

It is not a new router and is not intended to replace dsh-router-standard.

Behavior

The plugin only targets:

  • DeepSeek Harness 0.1.0-rc.7
  • the official standard preset
  • top-level sessions
  • DeepSeek V4 model IDs such as deepseek-v4-flash and deepseek-v4-pro

Before the first durable tool/call:

system:
You are a helpful software engineer assistant.

tools:
bash / pwsh
str_replace_editor

It also temporarily removes runtime contexts and hides automatic AGENTS.md / Skill Catalog injections.

After the first durable tool call, it restores the original Standard prompt, runtime contexts, full tool catalog, Skill Catalog, skill loader, and AGENTS.md injection.

The first promoted request also receives a one-shot capability re-routing reminder.

Evidence boundary

The upstream experiments provide direct evidence that the RL-shaped first-request surface can materially change DeepSeek V4 reasoning/action trajectories. They do not prove that the same trajectory always produces better final artifact quality.

dsh-router-standard PR #29 provides real-session evidence for hiding agent-instructions and skill-catalog during bootstrap and allowing them to return after promotion.

The one-shot promotion transition cue is experimental: it is motivated by a real post-promotion Skill-routing miss, but the cue itself has not been independently benchmarked.

Non-goals

This plugin intentionally does not include:

  • task classifiers
  • spec / react / weak routing
  • PTC / Code mode
  • We-Team Protocol
  • persistent near-field guidance
  • mode-boost
  • subagent wrappers
  • UI
  • hot reload infrastructure
  • custom presets

Non-V4 models, other presets, and subagent sessions are left untouched when the model is selected before the first request.

> v0.1.0 boundary: select the model before the first request. DSH rc.7 does not expose the Web session-local picker ref to ordinary plugins, so this release does not guarantee identical filtering/no-op semantics for model switches made while bootstrap is still awaiting promotion.

Install

Recommended:

dsh plugin --profile web add dsh-v4-anchor
dsh web

Or install the pinned GitHub release:

dsh plugin --profile web add github:Starfie1d1272/dsh-v4-anchor#v0.1.0
dsh web

To inspect the composed profile:

dsh --profile web --dump-config

Tests

No LLM/API calls are made.

npm run check
npm run pack:check

Provenance

See [THIRD_PARTY_NOTICES.md](./THIRD_PARTY_NOTICES.md).

License

MIT

This is an independent community project and is not an official DeepSeek project.

Back to searchOpen GitHub repository

More related plugins

View category