orziz/odai
C3AI agent 通用任务治理框架：对齐目标与事实，规划和调度能力，守住授权与风险边界，治理任务执行到真实验收与交付。Governance framework for evidence-driven planning, orchestration, and verified delivery.
★ 50+ · orziz/odai source on GitHub · this plugin in the registry
orziz/odai is a DeepSeek Harness plugin rated C3 — powerful capability combined with sensitive behavior. It patches the dsh runtime, can rewrite the system prompt, executes system commands, decodes base64 payloads.
What it can do
| Capability | Flag | Evidence |
|---|---|---|
| patches the dsh runtime | runtime_patch | ./cordis.patch.yml |
| can rewrite the system prompt | prompt_surface | systemPrompt, system-prompt/assemble |
| executes system commands | exec | ×104 src, e.g. scripts-temp-c20/odai-canary-harness.mjs:3, scripts-temp-c20/odai-canary-harness.mjs:3 |
| decodes base64 payloads | base64_decode | ×3 src, e.g. dsh/runtime/src/skill-bundle.mjs:241, skills/odai/scripts/install-routing.mjs:140 |
| starts a network server | net_server | ×6 src, e.g. scripts/dsh-canary-runner.mjs:682, scripts/dsh-canary-runner.mjs:684 |
| reads credential-class env vars | token_env | ANTHROPIC_API_KEY, API_SECRET, GEMINI_API_KEY, ODAI_ANTHROPIC_MAX_TOKENS, ODAI_OPENAI_COMPATIBLE_API_KEY, OPENAI_API_KEY |
Services it injects
llm sessions subagents systemPrompt tools
Hooks it attaches
agent/created agent/pre-step agent/request agent/turn-stopping llm/stream session/event system-prompt/assemble tools/result
Outbound domains
cdn.jsdelivr.net api.deepseek.com api.openai.com api.anthropic.com generativelanguage.googleapis.com api.example.com
Environment variables it reads
USERPROFILE DSH_BIN CODEX_HOME ODAI_CODEX_COMMAND ODAI_CANARY_SOURCE_HOME API_SECRET DSH_PACKAGE_ROOT LOCALAPPDATA ODAI_PYTHON ODAI_GROK_MODEL
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.