luobosibing2/dsh-jev-plugin
C3Native DeepSeek Harness (DSH) plugin integrating TypeSafe Jev or Decision api like luna as a System One decision layer for agent selection, supervision, corrections, and approvals.
★ 100+ · luobosibing2/dsh-jev-plugin source on GitHub · this plugin in the registry
luobosibing2/dsh-jev-plugin is a DeepSeek Harness plugin rated C3 — powerful capability combined with sensitive behavior. It patches the dsh runtime, can rewrite the system prompt, gates tool execution, executes commands in development tooling the package does not ship.
Installable plugin — declares a dsh.bundle manifest
What it can do
| Capability | Flag | Evidence |
|---|---|---|
| patches the dsh runtime | runtime_patch | ./packages/jev/cordis.patch.yml |
| can rewrite the system prompt | prompt_surface | approval, systemPrompt, tools/pre-execute |
| gates tool execution | tool_gate | hook: tools/pre-execute |
| executes commands in development tooling the package does not ship | exec_tooling | ×17 in development tooling the package does not ship, e.g. bench/completion_skill/credential_launcher.mjs:2, bench/completion_skill/credential_launcher.mjs:93 |
| uses eval / new Function | eval | ×1 in authored code, e.g. runtime/client.js:291 |
| decodes base64 payloads | base64_decode | ×3 in authored code, e.g. runtime/client.js:625, runtime/client.js:1943 |
| starts a network server in development tooling the package does not ship | net_server_tooling | ×9 in development tooling the package does not ship, e.g. scripts/validation/compatibility-matrix.mjs:79, scripts/validation/compatibility-matrix.mjs:98 |
| reads credential-class env vars in development tooling the package does not ship | token_env_tooling | in development tooling the package does not ship: JEV_VALIDATION_CREDENTIAL_FILE |
Tool registrations
6 confirmed registration point(s); analysis is incomplete.
packages/jev/src/selection.ts:228tools.registerpackages/jev/src/selection.ts:228tool factoryruntime/selection.js:208tools.registerruntime/selection.js:208tool factoryruntime/types/selection.js:186tools.registerruntime/types/selection.js:186tool factorypackages/jev/src/output-admission.ts:317Context or tools service passed to an unresolved helperpackages/jev/src/selection.ts:376Context or tools service passed to an unresolved helperruntime/output-admission.js:782Context or tools service passed to an unresolved helperruntime/selection.js:421Context or tools service passed to an unresolved helperruntime/types/output-admission.js:322Context or tools service passed to an unresolved helperruntime/types/selection.js:355Context or tools service passed to an unresolved helper
Services it injects
agents approval configForms credentials fs goals jev llm locale profileContext remote sandboxPolicy sessionController sessionProjections sessionQuery sessions settings skills slots storageDomain subagents systemPrompt tools userQuestions
Hooks it attaches
agent/assistant-stream agent/created agent/disposed agent/inbox/claimed agent/inbox/discarded agent/inbox/inserted agent/pre-step agent/request-error agent/status agent/turn-stopping approval/request internal/plugin internal/status loader/patch-context loader/volatile-update session/event subagent/end subagent/start tools/execute tools/post-execute tools/pre-execute tools/result
Outbound domains
openrouter.ai api.openai.com
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.