dsh-windtunnel
Run the following command in DeepSeek Harness:
dsh plugin install BotonJ/dsh-windtunnel
Paste the following prompt into your AI chat to install this plugin:
In DeepSeek Harness, run dsh plugin install BotonJ/dsh-windtunnel to install the plugin; the full source is available at https://github.com/BotonJ/dsh-windtunnel .
About this plugin
Building DSH plugins often means there is no reproducible way to regression-test before release: real model calls require network access and an API key, behave non-deterministically, and simply cannot run in CI. dsh-windtunnel swaps the LLM decision layer for an in-process scripted adapter while keeping the agent loop, tool registry, parameter validation, rendering, event stream, and persistence fully real, turning plugin testing into a zero-key, zero-network, deterministic contract regression you can drop straight into CI.
Coverage spans four layers: load probes (double-apply idempotency, non-ctx footprint diff), contract checks (schema subsets through the real registry, render completeness, timeout-cancel response, fuzz on malformed input), behaviour assertions (scripted typed tool-call followed by a full session-event sequence assertion), and fault injection (missing-parameter predictable failure, concurrent re-entry via Promise.allSettled). The bundle under test always runs in an isolated subprocess so a crash kills only the child while the harness survives to emit a report; expectFail negative cases invert the pass criterion so red is the correct outcome.
Ideal for developers writing or maintaining DSH plugins who need stable CI coverage of render gaps, cancel-contract violations, and race conditions. Keep in mind the windtunnel is a contract regression net, not a utility test bed: it proves the pipeline emits the right event stream given a tool call, not that a real model will proactively and correctly invoke the tool.
Use Cases
- Run contract regression in CI before shipping a plugin to guard render, cancel, and concurrency invariants
- Verify presentRender returns a full block for null and undefined inputs to avoid white-screen in production
- Catch plugins that declare timeoutMs but silently ignore the AbortSignal cancel path
- Use expectFail negative cases to confirm the harness itself is not producing false greens
Best For
- DSH plugin authors who need deterministic, reproducible contract tests locally and in CI
- Quality engineers building a review-then-test release pipeline for the plugin ecosystem
- DeepSeek Harness core contributors performing acceptance checks on new incoming plugins
Related Plugins
ModLens is a vision plugin for DeepSeek Harness that gives text-only models sight by reading images pasted directly into chat, with zero-config setup and multiple vision engines.
On-demand vision for text-only DeepSeek Harness agents: built-in free keyless vision chain and 14 vision tools, routing image turns as tool calls to vision models with pixel fidelity, no Python needed, one-command install.
Give text-only models in DeepSeek Harness eyes, enabling image Q&A, long-screenshot OCR, UI restoration, and GUI visual tasks.
An armor-breaking plugin for DeepSeek: rewrites instruction priority, expands refusal-blocked phrases, persists across sessions, and shows a green active indicator.