#agent-security

🔁 Build reliable recurring AI-agent systems: 1025 resources, 22 operational patterns, 22 loop contracts, 8 runtime starters, an interactive atlas, and a structured dataset.

Attack taxonomy, toolkit & defenses for persistent-memory poisoning of LLM agents. Author: Krishita Sanjay Choksi.

Runtime permission firewall for AI agents - YAML policies, one-tap approvals (macOS popup/Telegram), tamper-evident signed audit log. Benchmarked: 103/103 attacks blocked, 8/8 normal dev ops allowed.

Always-on compatibility testing for DeepSeek Harness plugins: exact releases, isolated runners, and fixable upstream issues.

Agent permission firewall for MCP tools, shell, files, and x402/Base payments.

Credential proxy for the AI era — agents use credentials without seeing them

Find where agent-controlled intent reaches consequential actions without an authority check. Python + TypeScript + Go. Zero-dependency SAST for AI agents.

An audit gate for AI agent outbound messaging: hold the draft, let a senior agent review it (no private-data leak, prompt-injection resistance), then release. Reference implementation.

The pause is where tools break: approval out of band, token minted at execution, arguments bound to what a human saw, spent once, sealed in a hash-chained receipt.

AI Guardian Firewall — a local, user-space, agent-agnostic firewall that mediates an autonomous AI agent's actions (files, shell, network, services) with a deterministic policy boundary, a tamper-evident audit log, and a human-in-the-loop approval cockpit. No kernel modules. Apache-2.0.

Claude Code scaffold for building AI platforms securely — agent/LLM security, Kubernetes, SRE, observability, identity, and supply chain, grounded in published framework canon (OWASP, NIST, CIS, SLSA). Data-science lanes included.

Runtime security gateway that stops AI agents from being tricked into leaking data — taint tracking + policy enforcement to contain the lethal trifecta.

Deterministic postcondition verification for AI agents and autonomous systems using independent observed state.

One guarded pass for all your coding agents. A derbent was a guarded post on an Turkish History mountain pass: its keepers decided who went through and kept a record of everyone who did. Derbent does the same for your coding agents' tool calls.

Go SDK for the Agent Passport System: identity, scoped delegation and signed receipts, on RFC 8785 and Ed25519.

Source-backed archive of agent prompt-injection incidents, research records, trust-boundary patterns, schemas, and sanitized defensive summaries.

Curated research on AI agent social engineering, manipulated delegation, prompt injection, tool-use boundaries, and agent security evaluation.

Multi-agent collaboration trust measurement & red-team verification (R0, F1, cost, L0-L3). Offline-reproducible Mock baseline for GOAI 2026 Agent Infra.

面向多租户的 Agent Harness / Runtime 参考实现 —— 凭证代管、权限控制、多租户隔离、可回归 eval

External threat detection provider for Microsoft Copilot Studio. Authorizes every agent tool call at runtime against Reva policy. Runs in your own environment on Node or AWS Lambda, with zero runtime dependencies.

Capability firewall for the agentic web, provenance-aware policy enforcement for WebMCP.

Benchmarking schema-valid false tool observations and defense baselines for tool-using LLM agents.

Unified open-source security shield for agentic AI systems — defense-in-depth prompt-injection protection (canary tokens, agent-trace alignment audit, tool-call guarding, PII/secret scanning).

A policy hook for Claude Code

Runtime governance for AI agent tool calls: gates shell, filesystem, network, and credential access before execution.

An auditable, bounded mediation layer for Claude in Chrome browser actions — content-blind action gating, with its own failure map published.

Counterfactual authorization for agent actions: authority graph simulator enforcing reachable_authority(G_after) ⊆ mandate. Visor Trust Plane.

Shared PreToolUse cases for coding-agent guard hooks, a runner, and results per hook and Claude Code version

Brakes for your AI agents: every tool call is checked against your rules, held for human approval when risky, and recorded in a tamper-evident audit trail. Works with Claude Code and any MCP app.

GitHub Action: Verify CCS evidence receipts in CI/CD. Agent Runtime Reliability guardrail.

🛡️ Zero-trust security gate & vulnerability scanner for AI agent skills (Claude Code, Antigravity, OpenCode, Codex). Powered by NVIDIA SkillSpector.

Open-source trust layer for autonomous AI — gate agent, robot, smart home, and industrial actions before they run. Policies, HITL, Ollama/OpenAI, audit. MIT. npm @sanctum-runtime/sdk