LeoGoat2004/dsh-eval-harness

C3

Diagnostic evaluation plugin for the DeepSeek Harness: confirmed harness/model/benchmark gates, hermetic traced episodes, and a standalone HTML report that attributes every failure to model, harness,

★ 0 · LeoGoat2004/dsh-eval-harness source on GitHub · this plugin in the registry

LeoGoat2004/dsh-eval-harness is a DeepSeek Harness plugin rated C3 — powerful capability combined with sensitive behavior. It patches the dsh runtime, runs code at install time, executes system commands, reads credential-class env vars.

Installable plugin — declares a dsh.bundle manifest

What it can do

CapabilityFlagEvidence
patches the dsh runtimeruntime_patch./cordis.patch.yml
runs code at install timeinstall_scriptpostinstall: npm run build
executes system commandsexec×53 in authored code, e.g. dsh-eval-harness/scripts/build.mjs:2, dsh-eval-harness/scripts/build.mjs:2
reads credential-class env varstoken_envSILICONFLOW_API_KEY

Tool registrations

8 static tool registration point(s).

Services it injects

tools

Outbound domains

pypi.tuna.tsinghua.edu.cn

Environment variables it reads

SILICONFLOW_API_KEY DSH_EVAL_DOCKER_TESTS DSH_EVAL_PIP_INDEX DSH_EVAL_KEEP_FIXTURE npm_execpath

How to read this

Levels measure capability surface and transparency, not maliciousness. A C3 plugin can be entirely legitimate — a desktop shell genuinely needs subprocesses. The point is that you can see this before installing. See the levels explained and how dsh plugins work.

Findings come from static analysis of shipped code; nothing is executed. Think a flag is wrong? Open an issue — every flag cites the file and line it came from.