Topic
Source-linked workflows that name security. Part of the AI agent use cases on UsecaseBoard, or browse all use cases.
Official CodeBuddy MCP docs show how to add local or remote MCP servers with codebuddy mcp add, pick user/project/local scopes, approve project servers, and lock tools with mcp__server__tool allow/deny rules.
Coder blog explains why local Claude Code is risky for enterprises, then shows AWS/EKS reference setup, golden templates with Claude Code registry module, managed settings, and audited sessions.
DiMilan shares a standing instruction: have your main Grok Bot run a prompt-injection safety check across every bot that can read email, calendar, files, or web pages before those connectors stay live.
An Indie Hackers founder almost shipped a no-auth “delete all users” endpoint Claude cheerfully wrote for a supposed test env—then built Thorim, a one-command layer of specialist agents/skills that makes coding agents challenge unsafe requests.
Cary Scardina’s Grok Bot checks headers, links, and attachment hashes with open-source tools and pre-scans domains/downloads so humans and downstream agents never open the low-hanging phishing bait first.
Brett had Grok Bot and Grok complete credit freezes at all three bureaus, an IRS protection PIN, data-broker opt-outs, 24/7 security alert watching, a daily news brief, bill checks, and family calendar dates in a single evening.
Anvith’s Day 8 Grok Bot challenge: stall a fake PayPal scammer, compare shoe prices across sites for the cheapest pair, and fill refund forms plus follow-ups he’d been putting off.
Security-minded user runs a Grok Bot that scans inbound email for phishing and social-engineering signals, checks headers, links, and attachment hashes against open-source tools, and pre-screens domains/downloads before humans or other bots open them.
Community JEV workflow: ask Grok Bot to create a JEV LAB folder on its cloud computer, install the JEV skill so an AI bodyguard checks risky commands and catches hidden instructions before the bot acts.
Two spoofed 'Netflix — payment on hold' emails hit inbox (hidden chars in sender name, reversed subject); Muse caught them, reported both to Microsoft's phishing desk, and deleted them.
MIT Python package signs every policy-gated tool call into a hash-chained .tkb file verifiable offline (Python+TS); works across Claude Code, OpenAI Agents, LangChain/LangGraph, MCP and more—agents can edit their own logs but not hide edits when the signer is out of reach.
Replit Bring Your Own Cloud (Oct 16) deploys cellular Replit—including the Agent harness—into enterprise Azure/GCP/AWS tenants: Full BYOC (private preview) for platform+apps in-tenant, BYOC for Published Apps (early access) for publish destinations with CMEK/tunnels, and Image Exports (Docker+Terraform+SBOM) for air-gapped self-host.
Local-first Apache-2.0 CLI for Claude Code/Codex/Cursor: mcp-xray ranks bloated tool schemas, sync compresses CLAUDE.md↔AGENTS.md↔Cursor rules, and shield fails CI on leaked keys/prompt-injection/Trojan Unicode—claims large cuts in model-visible prompt size without API calls.
Installable Claude Code pack (/launch-check) runs 440 read-only checks across security, payments, DB, SEO and AI tool permissions, then scores the app and lists file:line fixes; optional /launch-fix applies worst-first patches you review.
PATH-fronted proxies for terraform/aws/kubectl: TOML policies allow plan/read but block destroy/terminate/recursive deletes before the real binary runs—no LLM judgment in the safety path.
Perplexity open-sourced SPACE (Sandboxed Platform for Agentic Code Execution)—the Firecracker microVM sandbox behind Computer—under Apache 2.0 on Oct 15, 2026: HTTP/SSE APIs to create Linux environments, run commands, and manage lifecycle; partners CoreWeave, Akamai, and Crusoe plan to offer it; Semafor reports 0/118 breakout attempts vs other sandboxes in Perplexity’s tests.
Assign default-branch CodeQL/code-scanning alerts to Copilot: the agent explores the codebase, proposes a fix, reruns CodeQL to validate, iterates if needed, then opens a draft PR—batch up to 25 alerts or assign via REST with assignees ["copilot-swe-agent[bot]"].
AI Scan is GA: it complements CodeQL by auto-scanning PR diffs for languages/frameworks CodeQL does not cover (PHP, Shell, Terraform HCL, Dockerfiles, JSP, Blazor, etc.), surfaces advisory findings with explanations and suggested autofix, and consumes Code Security AI Credits without blocking merges.
Show HN: Abyss runs Pi/Hermes/Claude/Codex (any ACP agent) in Docker with path-aligned mounts, then applies host WASM plugins that can redact secrets, hydrate RAG, inject MCP, or auto-cancel sessions—agent-agnostic harness features without rewriting each agent plugin.