AI Agent Hub
Back to plugins
🧩

dsh-governance

admin-security Updated 2026.08.18

Run the following command in DeepSeek Harness:

dsh plugin install tappass/dsh-governance

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

Run dsh plugin install tappass/dsh-governance in DeepSeek Harness to install the plugin, available at https://github.com/tappass/dsh-governance, then set your TapPass API key and choose observe or enforce mode to activate governance.

About this plugin

Most AI safety guardrails answer "is this output harmful?"—a property of the model. The harder question is business-level: under our rules, right now, is this agent allowed to do this? dsh-governance shifts DeepSeek Harness from model safety into business authority governance, routing every tool call to the TapPass policy decision point where organisation-defined rules return an allow, deny, or escalate-for-human-approval verdict.

The plugin ships in observe mode by default: it watches and records every call without blocking anything, so you can size your policy against real traffic before enforcing a single deny. Rules are written in business language—"refunds over 500 need a human", "no customer PII leaves the EU region"—and the same set governs across harnesses and models without reimplementation.

It is aimed at teams deploying agentic AI who need auditable, business-defined authority controls, particularly those operating under the EU AI Act or spanning multiple toolchains. The plugin is deliberately thin—a bridge to /v1/govern—so harness API shifts require a small shim, not a rewrite.

Use Cases

  • Intercept sensitive tool calls (refunds, data egress) before execution and route to a policy engine for verdicts
  • Start in observe mode to map agent behavior, then gradually enable enforcement on critical operations
  • Apply the same business rules across multiple harnesses and models without per-platform reimplementation

Best For

  • Teams deploying agentic AI that require auditable, business-defined authority controls
  • Organisations under EU AI Act obligations needing unified AI behavior compliance across toolchains
  • Engineering teams running agents on multiple harness platforms seeking a single governance policy