AI Agent Hub
Back to plugins
⚙️

agpa-dsh-plugin

Workflow Updated 2026.09.15

Run the following command in DeepSeek Harness:

dsh plugin install eiainano/agpa-dsh-plugin

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install eiainano/agpa-dsh-plugin in your DeepSeek Harness terminal to install the plugin; the full source is available at https://github.com/eiainano/agpa-dsh-plugin.

About this plugin

Working across multiple Agent environments, from Claude Code to a custom harness, means your AGPA (Agent Player Achievements) progress should follow you everywhere. agpa-dsh-plugin fills the missing dsh-side bridge: it plugs into DeepSeek Harness as a Cordis plugin with a single design rule. DSH reimplements zero achievement logic; all seven achievement_* tools are thin shells that forward their arguments verbatim to the published AGPA MCP server, so the engine, achievement definitions, profiles, and the ~/.agent-achievements data store are shared across Claude Code, DSH, and any other harness. Cross-agent achievements fire automatically without duplication.

Core capabilities include a zero-runtime-dependency MCP-over-stdio bridge (Node builtins only, lazily spawning a single AGPA process), automatic event tracking that subscribes to dsh session/event and normalises tool/call and tool/result into CC-style payloads tagged with source dsh, each spawning one detached agpa-hook auto ingest process so cold-start downloads never lose events, and a SKILL.md that teaches the model when to track, poll, and announce unlocks. The full pipeline has been verified end-to-end on a real dsh 0.1.2-rc.1 build: a headless task successfully called achievement_stats and received live data from the shared store.

This plugin is for developers who run DeepSeek Harness alongside other Agent harnesses and want unified, cross-platform achievement tracking without maintaining duplicate logic. If you prefer a pure-bridge approach where the dsh side remains an inert thin shell and all evolution happens in the AGPA engine itself, this is the layer you need.

Use Cases

  • Trigger and query AGPA achievements inside DeepSeek Harness while sharing the same progress store as Claude Code
  • Automatically log tool-call events from headless or web sessions without manually invoking achievement APIs
  • Inspect cross-agent XP, levels, and unlocks from a single data layer instead of maintaining per-harness achievement state

Best For

  • Developers running DeepSeek Harness alongside Claude Code and other harnesses who want unified, cross-platform achievement data
  • dsh users who prefer a zero-duplication thin-shell bridge and want all evolution to happen on the AGPA engine side
  • Architects building multi-agent workflows that need standardised event tracking and achievement reporting across harnesses