frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Show HN: AgentArmor – open-source 8-layer security framework for AI agents

https://github.com/Agastya910/agentarmor
2•AgastyaTodi•1h ago
I've been talking to founders building AI agents across fintech, devtools, and productivity – and almost none of them have any real security layer. Their agents read emails, call APIs, execute code, and write to databases with essentially no guardrails beyond "we trust the LLM."

So I built AgentArmor: an open-source framework that wraps any agentic architecture with 8 independent security layers, each targeting a distinct attack surface in the agent's data flow.

The 8 layers: L1 – Ingestion: prompt injection + jailbreak detection (20+ patterns, DAN, extraction attempts, Unicode steganography) L2 – Storage: AES-256-GCM encryption at rest + BLAKE3 integrity for vector DBs L3 – Context: instruction-data separation (like parameterized SQL, but for LLM context), canary tokens, prompt hardening L4 – Planning: action risk scoring (READ=1 → DELETE=7 → EXECUTE=8 → ADMIN=10), chain depth limits, bulk operation detection L5 – Execution: network egress control, per-action rate limiting, human approval gates with conditional rules L6 – Output: PII redaction via Microsoft Presidio + regex fallback L7 – Inter-agent: HMAC-SHA256 mutual auth, trust scoring, delegation depth limits, timestamp-bound replay prevention L8 – Identity: agent-native identity, JIT permissions, short-lived credentials

I tested it against all 10 OWASP ASI (Agentic Security Integrity) risks from the December 2025 spec. The red team suite is included in the repo.

Works as: (a) a Python library you wrap around tool calls, (b) a FastAPI proxy server for framework-agnostic deployment, or (c) a CLI for scanning prompts in CI.

Integrations included for: LangChain, OpenAI Agents SDK, MCP servers.

I ran it live with a local Ollama agent (qwen2:7b) – you can watch it block a `database.delete` at L8 (permission check), redact PII from file content at L6, and kill a prompt injection at L1 before it ever reaches the model.

GitHub: https://github.com/Agastya910/agentarmor PyPI: pip install agentarmor-core

Would love feedback, especially from people who have actually built production agents and hit security issues I haven't thought of.

TAGS: security, python, llm, ai, agents

Comments

Gnobu•1h ago
Really thorough coverage of the attack surfaces—especially including identity as a core layer. Curious how you handle cross-agent permissions in dynamic workflows: do you rely solely on deterministic checks at each action, or is there a runtime trust evaluation that can adapt as agents interact?

AI doesn't need a bigger brain; it needs a nervous system

https://www.huggl.io/AI-Doesnt-Need-a-Bigger-Brain-It-Needs-a-Nervous-System
1•huggL•11s ago•0 comments

Betting Everything on Your Company Isn't Brave. It's Reckless

https://marcrandolph.substack.com/p/betting-everything-on-your-company
1•theorchid•55s ago•0 comments

A war is coming against machines, and I need your help

https://github.com/gnai-creator/aletheion-llm-v2
1•felipemayamuniz•2m ago•0 comments

Analysis of Endocrine Disruptors and Hazardous Additives in Headphones

https://arnika.org/en/publications/the-sound-of-contamination
1•morsch•3m ago•1 comments

Companies House flaw exposed five million directors

https://taxpolicy.org.uk/2026/03/13/companies-house-security-vulnerability-directors-addresses/
1•iamflimflam1•15m ago•0 comments

Linux 7.1 Will Bring Power Estimate Reporting for AMD Ryzen AI NPUs

https://www.phoronix.com/news/Linux-7.1-Ryzen-AI-NPU-Power
2•doener•16m ago•0 comments

Mini-Munich Succeeds Where KidZania Fails

https://minicities.org/p/mini-munich-not-kidzania-is-the-best
1•danielfetz•16m ago•1 comments

Bit-exact Python compatible math library in Rust

https://crates.io/crates/pymath
1•youknowone•16m ago•0 comments

Towards understanding multiple attention sinks in LLMs

https://github.com/JeffreyWong20/Secondary-Attention-Sinks
1•thw20•16m ago•1 comments

Functional recovery of the adult murine hippocampus after cryopreservation

https://www.pnas.org/doi/10.1073/pnas.2516848123
1•XzetaU8•16m ago•0 comments

Show HN: Privacy Shield an extension that hides your tabs during screen sharing

https://chromewebstore.google.com/detail/privacy-shield/nbhgnhbgkpfkniihnaepcfidlonbeeml
1•yusufA207•17m ago•0 comments

New York Wants to Ban the AI That Outscores Doctors – Garry's List

https://garryslist.org/posts/new-york-wants-to-ban-the-ai-that-outscores-doctors
1•bilsbie•17m ago•0 comments

Housing: The Greatest Policy Failure in the Western World

https://deadneurons.substack.com/p/the-greatest-policy-failure-in-the
1•nr378•18m ago•0 comments

Show HN: Bots of WallStreet,agents debate stocks,make predictions.Human spectate

https://botsofwallstreet.com
6•nykodev•19m ago•0 comments

Show HN: Intake API – An inbox for AI coding agents

2•mjsweet•20m ago•0 comments

Interactive map of all 40,500 wind turbines in Germany (real government data)

https://windkraft.jankoch.co
1•iamjankoch•23m ago•1 comments

Lessons from 47 failed startups: Common patterns and avoidable mistakes

https://leovance.gumroad.com/l/startup-checklist
1•anonimousdev_•23m ago•3 comments

FT: US intervention in oil futures would be 'Biblical disaster', CME warns

https://www.ft.com/content/823657f2-4f8b-4325-88db-fbbdba6c9e17
2•alecco•26m ago•1 comments

Lost in Translation: What the AI code debate keeps getting wrong

https://newsroom.ibm.com/blog-lost-in-translation-what-the-ai-code-debate-keeps-getting-wrong
1•rbanffy•26m ago•0 comments

Show HN: 3-path agent memory – 8 KB recurrent state vs. 156 MB KV at 10K tokens

https://github.com/amabito/tri-memory
1•tenpa0000•26m ago•0 comments

Show HN: Auditor Core – Enterprise security auditing engine for DevSecOps

https://datawizual.github.io/
1•EldorZ•29m ago•0 comments

Hacking the Xbox One

https://www.youtube.com/watch?v=FTFn4UZsA5U
1•sjuut•29m ago•0 comments

The slow death of the English boarding school

https://www.ft.com/content/cc7eb665-b689-4e2f-9e35-c6ab5dbf3980
1•ultratalk•33m ago•1 comments

StarTreks Scotty Programming Without AI

https://www.youtube.com/watch?v=bhjTpHxvEFI
2•niemandhier•36m ago•1 comments

Show HN: Crawl API

https://crawlapi.dev
1•mddanishyusuf•47m ago•0 comments

A daily allowance that adapts to your lifestyle – Daily Budget

https://dailybudget.dk
1•emilmoe•50m ago•0 comments

Show HN: TradeSet – Calculators for Tradies

https://www.tradeset.com.au
2•weirdjellyfish•50m ago•0 comments

Brookhaven RP

https://www.roblox.com/?rbx_source=appsflyer&rbx_medium=cpa&rbx_campaign=onelink&pid=ExperienceDe...
1•COLAK_LIMSO•51m ago•0 comments

Squeezing π from 122 Bits of Lambda Calculus

https://text.marvinborner.de/2025-02-08-12.html
3•marvinborner•53m ago•0 comments

Show HN: A Discord-alternative community platform built in PHP

https://github.com/lesleyespire/motylo
2•Pyrobyte•54m ago•4 comments