SpaceXAI 列出 56 个可交给 Grok Bot 的岗位

kozh ./ @mr_kozh原推发布于 2026.10.20

在 X 上打开原推

HOLY SH*T. SPACEXAI JUST LISTED 56 JOBS YOU CAN HAND TO GROK BOT. across those workflows, bots can work with inboxes, calendars, CRMs, expense data, vendor portals and engineering systems. that isn’t just automation. it’s a completely new security perimeter. 20 GitHub repos to stop your AI employee from becoming your next security incident: (1/20) Invariant Guardrails block dangerous sequences across messages and tool calls ▸ https://github.com/invariantlabs-ai/invariant (2/20) Agent Scan scan agents, MCP servers and skills for vulnerabilities ▸ https://github.com/snyk/agent-scan (3/20) Promptfoo red-team prompts and entire agent workflows inside CI ▸ https://github.com/promptfoo/promptfoo (4/20) Garak probe for jailbreaks, data leakage and unsafe behavior ▸ https://github.com/NVIDIA/garak (5/20) PyRIT automate adversarial testing against generative AI systems ▸ https://github.com/Azure/PyRIT (6/20) LLM Guard scan everything entering and leaving the model ▸ https://github.com/protectai/llm-guard (7/20) Guardrails AI reject malformed or unverified outputs before they reach another tool ▸ https://github.com/guardrails-ai/guardrails (8/20) E2B let agents execute code inside isolated cloud sandboxes ▸ https://github.com/e2b-dev/E2B (9/20) Daytona create disposable workspaces for AI-generated code ▸ https://github.com/daytonaio/daytona (10/20) Firecracker contain agent workloads inside lightweight microVMs ▸ https://github.com/firecracker-microvm/firecracker (11/20) gVisor add another isolation layer between agent code and the host ▸ https://github.com/google/gvisor (12/20) Open Policy Agent write explicit rules for which actions are allowed ▸ https://github.com/open-policy-agent/opa (13/20) Cerbos approve actions based on identity, resources and context ▸ https://github.com/cerbos/cerbos (14/20) OpenFGA define exactly who or which bot can access what ▸ https://github.com/openfga/openfga (15/20) Infisical give agents short-lived secrets instead of permanent credentials ▸ https://github.com/Infisical/infisical (16/20) OpenBao store, distribute and rotate sensitive keys ▸ https://github.com/openbao/openbao (17/20) OAuth2 Proxy put authentication in front of every internal tool ▸ https://github.com/oauth2-proxy/oauth2-proxy (18/20) Pomerium gate agent tools behind identity-aware, zero-trust access ▸ https://github.com/pomerium/pomerium (19/20) TruffleHog catch exposed credentials before your bot finds and reuses them ▸ https://github.com/trufflesecurity/trufflehog (20/20) Langfuse trace every prompt, tool call, cost and failure ▸ https://github.com/langfuse/langfuse none of these repositories is Grok Bot’s source code. they are the open-source security layer you need around any agent with real tools and production credentials. the stack: scan before connecting → red-team before deploying → isolate every execution → authorize every action → expose secrets only when needed → trace everything afterward an agent that can only write text is a chatbot. an agent that can open your inbox, change your CRM and execute code is an employee with credentials. secure it like one. save this before your always-on bot gets the keys to production ↓

这条是引用着另一条写的：x.com/mr_kozh/status/2109231794457608494