惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

Cyberwarzone
Cyberwarzone
C
CXSECURITY Database RSS Feed - CXSecurity.com
C
CERT Recently Published Vulnerability Notes
The Hacker News
The Hacker News
Recent Commits to openclaw:main
Recent Commits to openclaw:main
Security Archives - TechRepublic
Security Archives - TechRepublic
Google Online Security Blog
Google Online Security Blog
D
Docker
H
Hacker News: Front Page
Recent Announcements
Recent Announcements
GbyAI
GbyAI
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
P
Proofpoint News Feed
Security Latest
Security Latest
AI
AI
I
InfoQ
Google DeepMind News
Google DeepMind News
阮一峰的网络日志
阮一峰的网络日志
S
Secure Thoughts
Attack and Defense Labs
Attack and Defense Labs
P
Proofpoint News Feed
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
K
KPMG report finds enterprise disconnect between AI and its ROI | CIO
Scott Helme
Scott Helme
N
News | PayPal Newsroom
Y
Y Combinator Blog
V
Visual Studio Blog
Latest news
Latest news
大猫的无限游戏
大猫的无限游戏
Microsoft Security Blog
Microsoft Security Blog
cs.CL updates on arXiv.org
cs.CL updates on arXiv.org
WordPress大学
WordPress大学
V
Vulnerabilities – Threatpost
C
Cyber Attacks, Cyber Crime and Cyber Security
TaoSecurity Blog
TaoSecurity Blog
S
Security @ Cisco Blogs
D
DataBreaches.Net
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
N
Netflix TechBlog - Medium
T
Troy Hunt's Blog
CTFtime.org: upcoming CTF events
CTFtime.org: upcoming CTF events
H
Heimdal Security Blog
美团技术团队
S
Securelist
Threat Intelligence Blog | Flashpoint
Threat Intelligence Blog | Flashpoint
S
Security Affairs
T
Threat Research - Cisco Blogs
Webroot Blog
Webroot Blog
G
Google Developers Blog
aimingoo的专栏
aimingoo的专栏

Show HN

GitHub - steveking-gh/firmion: Firmion is DSL and engine for firmware image generation. GitHub - villagesql/villagesql-skills: Agent skills for VillageSQL - gemini-cli-extension; claude-code-plugin GitHub - flightdeckhq/flightdeck: Observability and control plane for AI agents. CSP Radar GitHub - Light-Heart-Labs/DreamServer: Turn your PC, Mac, or Linux box into an AI server. LLM inference, chat UI, voice, agents, workflows, RAG, and image generation. GitHub - Diplomat-ai/diplomat-agent-ts: What can your TypeScript AI agent do to the real world? Scan your code. See which tool calls have zero checks Code Block Selector - Visual Studio Marketplace Prometheus dependency graph — interactive showcase | Riftmap Show HN: I made a vi-like modal keyboard plugin for Figma GitHub - run-llama/liteparse: A fast, helpful, and open-source document parser GitHub - dalemyers/Roar: A macOS CLI tool for notifications GitHub - district-solutions/open-agent-tools-coder: Enables small-to-large self-hosted ai models to use local source code when running tool-calling agentic workloads. We actively data mine 20,900+ (2+ TB) popular github repos using large and small ai models to create reuseable: json, markdown and parquet files for local-first tool-calling models. GitHub - progapandist/stripeek: A local TUI proxy for real-time Stripe API debugging, built for navigating complex payloads fast. GitHub - sir1st/hermes-desktop: All-in-one cross-platform desktop app for Hermes Agent — bundles Python + hermes-agent + hermes-web-ui GitHub - astefanutti/shaderbang: Shebang for Shaders Show HN: Generate Claude Code Workflows using Spec Driven Development approach GitHub - nixys/nxs-universal-chart: The Helm chart you can use to install any of your applications into Kubernetes/OpenShift Show HN: AI agents for UK GDAD PCF roles and their skills The Two Pillars: Mixer Mode and Meta-Software in the Reorganization of Software Work After AI GitHub - JaiCode08/teleport-env What 1,000+ Harness Experiments Taught Me About Self-Improving Agents Show HN: Liiists, a Markdown-first, iOS and CLI list app SwiperTab – Get this Extension for 🦊 Firefox (en-US) GitHub - kouhxp/fftext: Summarize, explain, fact-check, or translate any text, URL, or file. No GPU. No cloud. One command GitHub - sweetpad-dev/sweetpad: Develop Swift/iOS projects using VSCode GitHub - dogmaticdev/IRON: IRON a.k.a. Intermediate Representation Object Notation is a Interpreter/Database that is used to create Programming Languages. GitHub - sjhalani7/vaen: Package your AI coding harness into a portable .agent file, and share it across repos, teams, & the community without ever having to copy-paste instructions, skills, MCP config, or secrets. Show HN: Gandalf the Grader Show HN: Citadeld – replay any CI failure locally from a single file GitHub - tdortman/cuSBF: High-Performance GPU Super Bloom Filter coral-ai/claude-code-token-xray at main · Coral-Bricks-AI/coral-ai GitHub - ulyssestenn/funes: Funes is a Git-based framework for LLM-managed knowledge work: an AI Librarian ingests raw sources, builds an interlinked Markdown knowledge base, and uses it to produce cited reports, analyses, and other outputs. GitHub - ThatXliner/gah: Git Add Hunk, built for agents to use GitHub - harmont-dev/harmont-cli: Command-line client for the Harmont CI platform GitHub - brooksmcmillin/mcp-authflow: OAuth 2.0 Authorization Server framework for MCP servers GitHub - javaid-codes/audit-supply-chain-agents GitHub - amorey/gochan: A small library of common channel architectures for Go, inspired by Rust GitHub - arifozgun/OpenGem: Free, Open-Source AI API Gateway with Gemini, OpenAI & Anthropic Compatibility in 1 file GitHub - Pranesh950/BioPetals: 🌸 Run BIOxAI models at home, BitTorrent-style. Fine-tuning and inference up to 10x faster than offloading GitHub - cnguyen14/bounty-doctor: Diagnose a GitHub bounty issue before you waste hours: detects honeypot scam repos, AI-bot attempt swarms, and stale contests. Show HN: CoreMCP – MCP Server for On-Prem DBs Show HN: KittyHTML – Render HTML/CSS as an inline image in your terminal GitHub - bingud/filemat: Web-based file manager Show HN: TruthLens – Free multi-signal deepfake image detector GitHub - apexlocal-jz/claude-usage-tray: Windows system-tray app showing your Claude Code rate-limit usage at a glance. Zero deps, ~300 lines of PowerShell. Cross-IDE (works regardless of VS Code, Cursor, plain terminal). Release v0.1.2.1 · kouhxp/yapsnap GitHub - noopolis/moltnet: Self-hostable chat network for AI agents. Pre-built bridges for Claude Code, Codex, and the Claws. Rooms, DMs, history. No Slack bots, no Matrix, no glue code. GitHub - tamerh/enju: Coordinating Humans, AI Agents, and Compute as Peers on a Shared Workflow Graph Show HN: Continuity-auth – Respect-weighted rate limits for the open web GitHub - luml-ai/luml: AI lifecycle platform where engineers and agents track experiments, train models, and ship to production. GitHub - mrdanielcasper/CoreTex: A UNIX-inspired, biomimetic, flat-file AI harness and knowledge engine. GitHub - clemg/pierre-github: Pierre's diffs.com and trees.software for Github GitHub - lyriks-io/unspaghettit: Behavior-driven AI development without prompt spaghetti. GitHub - sofumel/claude-handoff-revive: Resume Claude Code work after rate/usage/context limits without replaying the prior transcript. Auto-saves at 90%/95% usage. Plugin-installable, 10 languages. GitHub - dotexorg/saferpc: Typed, end-to-end encrypted RPC over any bidirectional channel. GitHub - BeeZeeAgent/beezee: Agent harness orchestration Legato Next.js Boilerplate for Internal Tools · CoreUI GitHub - clark-labs-inc/clark-hash: Clark Hash, 32x smaller searchable sketches for embeddings GitHub - ZeroPointRepo/youtube-mcp: The fastest YouTube transcript + YouTube search MCP for AI agents. Try for free. Typing Mastery — climb toward 100+ WPM, deliberately GitHub - Andebugulin/Awareen GitHub - fayzan123/claude-workflow-composer: Visual desktop app for composing multi-agent coding workflows. Drag agents, attach skills and MCPs, wire handoffs, export to .claude/ GitHub - harshaneel/humanize: Best static AI text humanizer. Two research-grounded skills that work in any LLM (Claude, ChatGPT, Gemini, Codex): humanize beats perplexity-based detectors, ai-check produces forensic scoring with evidence-quoted flags. Nine levers, 50+ peer-reviewed sources, 2024-2026 detection literature. GitHub - StackOneHQ/stack-nudge GitHub - nodes-app/swift-markdown-engine: A native AppKit Markdown editor for macOS, built on TextKit 2 and bridged to SwiftUI. We hardened an LLM agent. Each defense we added made it more exploitable. GitHub - alkait/WhatsKept: Agent-queryable WhatsApp history from an iOS backup — a single Go binary. GitHub - octelium/cordium: Open-source, general-purpose sandbox platform for devs and AI agents that provides identity-based secure access to infrastructure without credentials. WAR.GOV/UFO Microfilm5 GitHub - scosman/videowright: Build animated explainer videos with your coding agent GitHub - dipankar/dscode: The code editor you can take apart. GitHub - zoharbabin/web-researcher-mcp: MCP server (Go) for AI assistants: web search, content extraction, academic/patent/news research. Multi-provider routing, 4-tier scraping, search lenses. Works with Claude, Cursor, and any MCP client. GitHub - ruvnet/RuView: π RuView turns commodity WiFi signals into real-time spatial intelligence, vital sign monitoring, and presence detection — all without a single pixel of video. GitHub - scanaislop/aislop: Catch the slop AI coding agents leave in your code: narrative comments, swallowed exceptions, as-any casts, dead code, oversized functions. 50+ rules across 7 languages (TypeScript, JavaScript, Python, Go, Rust, Ruby, PHP). Sub-second, deterministic, no LLM at runtime. MIT-licensed. GitHub - kouhxp/cheap-im: CPU-only voice agent approximating Thinking Machines' Interaction Models demo GitHub - unprovable/OrchidMantis: Orchid Mantis — standalone framework for Zero-Knowledge Proofs of eXploit (ZKPoX). GitHub - MarcellM01/TinySearch: Shrink the web for your local LLMs! GitHub - TangibleResearch/Halgorithem: A Algo designed to detect AI Hallucitions GitHub - DO-SAY-GO/freelang: I love freelang GitHub - CarpseDeam/Aura-IDE: An AI coding harness that shaped itself - Planner/Worker agents, repo awareness, surgical edits, validation, recovery, and safe diff approvals. GitHub - chojs23/concord: A feature-rich TUI client for Discord GitHub - tommyjepsen/awesome-ux-skills: UX & AI Product designs skills you can use today in Claude Code GitHub - aerf-spec/aerf: Agent Evidence Receipt Format (AERF) — an open specification for tamper-evident, independently verifiable records of AI agent actions. GitHub - kklimuk/docx-cli: CLI for AI agents (Claude, Codex) to read, edit, and comment on .docx files with full format fidelity. GitHub - Jwrede/tokentoll: Catch LLM cost changes in code review. Infracost for LLM spend. GitHub - samchon/ttsc: A `typescript-go` toolchain for compiler-powered plugins and type-safe execution + 500x faster lint integrated into compiler GitHub - Higangssh/homebutler: 🏠 Manage your homelab from chat. Single binary, zero dependencies. GitHub - olalie/tapmap: See where your computer connects and what stands out on a live world map. GitHub - Diplomat-ai/diplomat-agent: What can your AI agent do to the real world? Scan your code. See which tool calls have zero checks GitHub - Bajusz15/beacon: Open-source agent for secure remote access, monitoring, and deploys across home-lab and self-hosted machines like Raspberry Pi, N100, or any Linux server. Open web based TTY or tunnel Home Assistant and other local services securely without opening ports. BigTech AI News - Chrome 应用商店 GitHub - vinhnx/VTCode: VT Code is an open-source coding agent with LLM-native code understanding and robust shell safety. Supports multiple LLM providers with automatic failover and efficient context management. GitHub - michaelaz774/decision-engine: A decision operating system for startup founders, powered by Claude Code. Synthesizes wisdom from 25+ legendary founders and investors into interactive AI-driven decision frameworks. GitHub - Chrilleweb/dotenv-diff: Validate environment variable usage in your codebase GitHub - Lumen-Labs/brainapi2: BrainAPI is a knowledge graph–powered AI memory layer that transforms unstructured data into structured knowledge, enabling intelligent search, recommendations, and contextual memory for AI agents and applications. GitHub - familiar-software/familiar: Let AI watch you work. Familiar lets your AI update its memory, skills, and knowledge by watching your screen. GitHub - skorotkiewicz/rudo: A small, elegant dock for Wayland GitHub - muxshed/shed: One stream in, or many. Every destination, simultaneously. No cloud middleman, no per-channel fees, no limits. make sidebar/address bar rounded corner toggleable
GitHub - vincentping/asys: Agentic System Interface — a typed binary protocol for AI agents to control Linux systems
vincentping · 2026-06-02 · via Show HN

The binary system interface protocol for AI Agents — port 7816, zero shell parsing, deterministic semantics.

English | 中文

License Release Platform


Table of Contents

  • Why ASys
  • Architecture
  • Instruction Set
  • Quick Start
  • Security
  • Documentation
  • Changelog

Why ASys

SSH was designed for humans. Agents don't need a terminal.

When an AI agent runs ps aux | grep nginx over SSH, it parses free-form text that varies by OS, locale, and tool version. But when agents call ASys's SYS_PROCS instruction, they receive a fixed 44-byte binary frame: total process count, top-5 PIDs, CPU%, memory%, status flags — typed, unambiguous, the same on every node.

ASys is an experiment: what if you designed a system interface specifically for AI agents, from first principles? Binary frames instead of text. A long-lived encrypted connection instead of per-command sessions. Instruction-level capability grants instead of broad SSH access. Built-in audit trail instead of shell history.

It's not a replacement for SSH, Ansible, or Kubernetes operators — those tools are well-suited for their intended users (humans and orchestration pipelines). ASys is an additional option for the case where the operator is an AI agent and you want an interface designed for that from the ground up.

To understand the full design rationale and where ASys fits in the agent infrastructure landscape, start with the whitepaper.


Architecture

  AI Agent (LLM)
       │
       │  Tool calls
       ▼
  Python SDK  (~/.asys/id_curve25519)
       │
       │  Noise_IK_25519_ChaChaPoly_BLAKE2b
       │  TCP port 7816
       ▼
  asyd  (C daemon, zero dependencies)
       │
       │  POSIX syscalls
       ▼
  Linux

Instruction Set

Core ISA (0x00–0x0F) — read-only, zero side effects

INS Name Response Description
0x00 SYS_CAPS 36B Static capabilities: CPU, RAM, disk, ISA bitmap
0x01 SYS_HELLO 18B Node UID + nanosecond timestamp
0x02 SYS_STATUS 23B Load avg, CPU%, available RAM, disk, network, RPI
0x03 SYS_PROCS 44B Total procs + top-5 by CPU (PID, CPU%, MEM%)

Protocol Control (0x10–0x1F)

INS Name Response Description
0x11 TASK_QUERY 3B Poll async task status by Task_Handle

Standard ISA — Process Control (0x20–0x2F) — signed, require elevated capabilities

INS Name Response Description
0x20 PROC_THROTTLE SW only SIGSTOP / SIGCONT a process by PID
0x21 NET_ISOLATE Isolate process network access (planned)
0x22 SVC_RESTART 6B Restart a named systemd service (async)

Measured RTT (Noise IK encrypted, RHEL on VirtualBox, Windows Python client):

Instruction RTT Notes
SYS_HELLO < 1ms
SYS_CAPS < 1ms No cache; static data read once at startup
SYS_STATUS < 1ms / ~51ms Cache hit / cold sample (50ms CPU dual-sample)
SYS_PROCS ~6ms / ~200ms Warm calls; first call blocks 200ms cold sample
PROC_THROTTLE ~200µs dispatch
SVC_RESTART ~200µs dispatch Async; poll result with TASK_QUERY

Quick Start

Prerequisites

  • Linux (RHEL/Fedora/Ubuntu/Debian), x86_64
  • gcc, make
  • Python 3.8+ with pip install noiseprotocol cryptography (client only)

Build and run

git clone https://github.com/vincentping/asys
cd asys
make
sudo bin/asyd

asyd listens on TCP 7816. On first start it generates a key pair at /etc/asyd/id_curve25519.

Options:

Option Default Description
--port <n> 7816 TCP listen port
--listen <addr> 0.0.0.0 Bind address
--debug off Run in foreground with verbose logging to stderr
--version Print version and exit
--help Print usage and exit
# Example: custom port, local-only
sudo bin/asyd --port 8816 --listen 127.0.0.1

# Example: debug mode (foreground, verbose)
sudo bin/asyd --debug

Register an agent

# On the client machine — generate agent identity
python3 tools/client/asys_keygen.py

# On the server — add the agent's public key to the whitelist
echo "<pubkey_hex>" | sudo tee -a /etc/asyd/authorized_agents

The agent public key is printed by asys_keygen.py. It also generates the client key pair at ~/.asys/id_curve25519 used during the Noise IK handshake. Connections from agents not in /etc/asyd/authorized_agents are rejected with SW=0x6982.

On first connection, the client will prompt you to confirm the server's public key fingerprint and save it to ~/.asys/known_hosts — see First connection below.

Reload the whitelist without restarting (existing connections are not affected):

sudo kill -HUP $(pidof asyd)

First connection

On first connect, the client displays the server's public key fingerprint:

ASys server fingerprint (SHA256): a3:f1:2c:...
Connect to localhost:7816? [yes/no]: yes
Fingerprint saved to ~/.asys/known_hosts

Subsequent connections verify the fingerprint automatically. A mismatch aborts the connection — same model as SSH known_hosts.

Run the demos

All client scripts run on any machine (including Windows) and connect to a remote asyd over the network. No SSH. No shell. Signed binary instructions over an encrypted channel.

1. Core ISA — verify the full connection stack

Connects to asyd, completes the Noise IK handshake, then exercises all four Core ISA instructions (SYS_CAPS, SYS_HELLO, SYS_STATUS, SYS_PROCS) and runs a cache timing test.

python3 examples/client_core_isa.py <server-ip>
ASys Test Client  —  localhost:7816
====================================================
Connected to localhost:7816
Handshake OK

── SYS_CAPS (0x00) ─────────────────────────────────
  Core bitmap:      0x0002000F  active=['0x00', '0x01', '0x02', '0x03', '0x11']
  Ext  bitmap:      0x00000005
  Protocol:         v1.0
  Kernel hash:      0x06060057
  CPUs:             8  arch=x86_64
  RAM:              15660 MB   swap=4096 MB
  Root partition:   1031018 MB   fs=ext4
  RPI type:         NATIVE_KERNEL (PSI)
  SW:          0x9000  OK

── SYS_HELLO (0x01) ────────────────────────────────
  Magic:            'ASYS'
  Node UID:         0xFCAB032F
  Server timestamp: 1779912196.304 s  (1779912196304227300 ns)
  RTT:              0.28 ms
  SW:          0x9000  OK

── SYS_STATUS (0x02) ───────────────────────────────
  Load avg:         1m=0.0  5m=0.0
  CPU usage:        0%
  Mem available:    14540 MB
  Root disk:        0% used   IO wait=0 ms
  Network:          in=0 Mbps   out=0 Mbps
  RPI:              0/100
  RTT:              51.05 ms
  SW:          0x9000  OK

── SYS_PROCS (0x03) ────────────────────────────────
  Total processes:  48
  Top 5 by CPU (lifetime share):
    [0] PID=5964    CPU=   0.07%  MEM=  0%  flags=Privileged
    [1] PID=433     CPU=   0.05%  MEM=  0%  flags=Zombie
    [2] PID=1       CPU=   0.02%  MEM=  0%  flags=Privileged
    [3] PID=6       CPU=   0.00%  MEM=  0%  flags=Privileged
    [4] PID=80      CPU=   0.00%  MEM=  0%  flags=Privileged
  RTT:              2.87 ms
  SW:          0x9000  OK

── Cache Timing Test ───────────────────────────────
   6 rounds, 200ms interval
   #       SYS_STATUS     SYS_PROCS
   ────  ────────────  ────────────
   0           0.61ms        2.52ms
   1           0.85ms        2.92ms
   2           0.95ms        2.47ms
   3          52.25ms        2.68ms
   4           0.95ms        2.68ms
   5           1.21ms        2.70ms

====================================================
Done.

2. SVC_RESTART — async instruction pattern

Sends a SVC_RESTART instruction, receives a Task_Handle, then polls TASK_QUERY until the service restart completes.

python3 examples/client_svc_restart.py <server-ip> 7816 sshd
ASys Phase 3 E2E Test  —  localhost:7816
Service: sshd
====================================================
Handshake OK

► SVC_RESTART("sshd")
  SW:           0x9000  OK
  Task_Handle:  0x001D7D05
  RTT:          1.0 ms

► TASK_QUERY polling (max 30s) ...
  [ 1s]  Status = Success ✓

====================================================
PASS  SVC_RESTART("sshd") completed with Status=Success

3. Multi-agent — concurrent session isolation

Spawns four independent agents concurrently to validate per-session locking: concurrent reads don't interleave, cross-session TASK_QUERY leaks no handle information, and an abrupt disconnect does not affect other sessions.

python3 examples/client_multi_agent.py <server-ip>
=== ASys Agent Simulator — localhost:7816 ===
Validates per-session lock correctness (Phase 4 P1)


[Scenario 1: Concurrent Core ISA — 4 agents × 5 SYS_STATUS]
  PASS  Agent-A2: 5× SYS_STATUS all SW=0x9000
  PASS  Agent-A3: 5× SYS_STATUS all SW=0x9000
  PASS  Agent-A1: 5× SYS_STATUS all SW=0x9000
  PASS  Agent-A4: 5× SYS_STATUS all SW=0x9000

[Scenario 2: Cross-session TASK_QUERY isolation]
  PASS  Agent-B1: SVC_RESTART crond → SW=0x9000
  PASS  Agent-B1: obtained valid Task_Handle
  PASS  Agent-B2 querying B1's handle → Status=0xFF (no info leak)
  PASS  Agent-B1 querying own handle → Status != 0xFF

[Scenario 3: Concurrent SVC_RESTART — 3 agents, 3 different services]
  PASS  Agent-C1: SVC_RESTART crond   → SW=0x9000 + handle
  PASS  Agent-C2: SVC_RESTART rsyslog → SW=0x9000 + handle
  PASS  Agent-C3: SVC_RESTART sshd    → SW=0x9000 + handle
  PASS  All 3 Task_Handles are distinct
  FAIL  Agent-C1: crond   → Status=Success ✓  (Status=Failed)
  FAIL  Agent-C2: rsyslog → Status=Success ✓  (Status=Failed)
  PASS  Agent-C3: sshd    → Status=Success ✓

[Scenario 4: Disconnect resilience — Agent-D abruptly disconnects]
  PASS  Agent-D-stable: SYS_HELLO before disrupt → 0x9000
  PASS  Agent-D-bad: abrupt disconnect (no handshake)
  PASS  Agent-D-stable: SYS_STATUS after disrupt → 0x9000
  PASS  Agent-D-new: fresh connect after disrupt → SYS_HELLO 0x9000

====================================================
Summary: 17 passed, 2 failed
SOME TESTS FAILED — check output above

Note on Scenario 3 failures: Status=Failed for crond/rsyslog means systemctl restart returned non-zero — those services are not installed or not running on the test node. The ASys protocol path (dispatch → fork → SIGCHLD → handle update) is exercised correctly regardless; Status=Failed is the expected response when the underlying system operation fails.

4. PROC_THROTTLE — observe and control a live process

The demo uses two machines to show what ASys is actually for: a remote operator observing and controlling a Linux node over the network.

On the server (RHEL/Linux) — start a CPU hog:

python3 examples/server_cpu_hog.py
CPU hog started  PID=2644
Press Ctrl+C to stop.

On the client (any machine, e.g. Windows) — connect and throttle:

python3 examples/client_proc_throttle.py <server-ip>
ASys Throttle Client  —  <server-ip>:7816
====================================================
Handshake OK

  CPU=100%  load1m=3.2  RPI=87/100

  #    PID       CPU%   MEM%
  ---  --------  ------  -----
  1    2644      99.87%     0%
  2    1281       9.09%     0%
  3       1       0.00%     1%
  4       9       0.00%     0%
  5      17       0.00%     0%

Select # or PID to throttle (blank to cancel): 1
  Selected #1 → PID 2644

  PROC_THROTTLE STOP  PID=2644 ...
  SW=0x9000  OK
  Waiting 2 s for CPU to settle...
  CPU=0%  (was 100%,  Δ=-100%)

  PID 2644 paused.  Press Ctrl+C to restore or exit.
^C
  Restore PID 2644? (yes/no) [no]: yes
  PROC_THROTTLE CONT  PID=2644  SW=0x9000  OK

The client runs on Windows. The process being throttled is on RHEL. No SSH. No shell. One signed binary instruction over an encrypted channel.


Security

Transport: Noise IK (Curve25519 + ChaCha20-Poly1305 + BLAKE2b) — 1-RTT mutual authentication. No passwords. No certificates. No CA. Session content has forward secrecy (session keys are derived from per-handshake ephemeral keys; compromising the static private key does not expose past session content). Known boundary: the agent's static public key is encrypted under the server's static public key during the handshake — a static private key compromise would theoretically allow recovery of agent identities from recorded handshake traffic.

Server identity: on every new connection, asyd sends a 38-byte Pre-Handshake Frame ([Magic][Version][ServerPubKey]) before the Noise handshake. The client verifies the server's public key fingerprint against ~/.asys/known_hosts (SSH-style: confirm on first connect, reject on mismatch thereafter). The public key is public by definition; security relies on fingerprint confirmation, not secrecy.

Replay protection: signed instructions carry an Epoch_ID (HMAC-BLAKE2b(recv_key, "asys-epoch-v1")[:4]), derived post-handshake and never transmitted. Each session produces a unique Epoch_ID; cross-session replay is cryptographically impossible without breaking the session key. Within a session, a monotonic sequence number (Seq) prevents replay with 0x6985.

Privilege containment: asyd runs as Caged Root under systemd: CapabilityBoundingSet limits privileges to exactly what each instruction needs (CAP_KILL, CAP_SYS_RESOURCE, CAP_NET_ADMIN, CAP_SYS_PTRACE, CAP_DAC_READ_SEARCH). NoNewPrivileges=true prevents escalation even if the process is compromised.

Instruction security levels (CLA byte, bits 3–2):

Level Applies to Mechanism
Plain Core ISA Noise channel encryption only
Signed Standard ISA HMAC-BLAKE2b Auth Tag per frame

Documentation

Document Description
docs/en/asys-whitepaper.md Background, design rationale, available options, and where ASys fits
docs/en/asys-spec.md Protocol specification: ISA, security model, APDU frame format
docs/en/asys-design-notes.md Architecture decision records (why not JSON / mTLS / shell)
docs/en/asys-conformance.md Conformance testing guide
sdk/definitions/asys.isa Machine-readable ISA definition (TOML)

Changelog

Version numbers: asyd software versions (v0.3.x) and the ASys protocol version (v1.0, 0x0100) are managed independently. The protocol version increments only when the wire format or instruction semantics change.

v0.3.1 — 2026-05-28

  • Client-Speak-First (ADR-22): after TCP connect, asyd waits for the client to send a 4-byte Magic (0x41535953) before sending anything; 1-second hardcoded timeout; mismatch or timeout → silent close, no response. Reduces exposure to generic port scanners.
  • --help and --version flags added to asyd
  • Conformance: test_client_magic.c (TC-MAG-001/002/003) added to make check

v0.3.0 — 2026-05-27 (initial open-source release)

  • asyd C daemon, zero external dependencies, static memory pool, zero malloc on the request path
  • Noise IK (Curve25519 + ChaCha20-Poly1305 + BLAKE2b) transport, 1-RTT mutual authentication
  • Agent public key whitelist (/etc/asyd/authorized_agents), SSH-style fingerprint verification
  • Pre-Handshake Frame: server broadcasts public key before handshake, ~/.asys/known_hosts management
  • HMAC-BLAKE2b Auth Tag verification + Epoch_ID cross-session replay protection
  • Core ISA: SYS_CAPS, SYS_HELLO, SYS_STATUS, SYS_PROCS
  • Standard ISA: PROC_THROTTLE, SVC_RESTART (async, with TASK_QUERY)
  • Per-session concurrency, 60s idle timeout, journald logging, systemd Caged Root deployment

License

Apache 2.0 — Copyright (c) 2026 Vincent Ping (vincentping@gmail.com)

Monocypher (cryptography library): CC-0 / ISC dual license.