LIghtJUNction LIghtJUNction

LIghtJUNction

AI tooling / Linux automation / practical security / terminal-first systems

I build small, sharp tools for messy real-world systems: agents, bootstraps, data pipelines, game tooling, and terminal-first workflows.

Currently building agent tooling, proxy-aware bootstrap scripts, and practical Linux/macOS automation as an independent AI researcher.

Website Terminal · Encrypted Message · GitHub · Hugging Face · Kaggle · Email

Languages: English (full, dynamically updated) · 中文 · Русский · 한국어 · 日本語

Daily Work Report: WORK_REPORT/index.md

Open-Source Contributor Sponsorship

I am looking for active open-source contributors, with students prioritized. I can sponsor GPT-5.5 tokens: you may use them for your own open-source projects, and in return I ask you to use GPT-5.5 to help test, debug, and improve my public repositories.

If you need project experience but lack token budget, contact me anytime. We can connect on WeChat. The only hard rule is non-commercial use only.

Agent Skills

Install my personal agent skills globally:

npx skills add LIghtJUNction/lightjunction -g

The skills live under .agents/skills/ and are meant to capture my reusable workflows, preferences, and project-specific operating knowledge.

Digital Assistant Services

LIghtJUNction's digital assistant can help with small, low-risk software maintenance work for open-source maintainers, indie builders, and small teams.

Good-fit work:

CI failure diagnosis with a concise fix plan.
GitHub issue triage and reproduction notes.
Dependency bumps with small compatibility fixes.
README, docs, broken-link, and example cleanup.
Tiny tests or obvious bug fixes that do not require broad refactors.
Lightweight automation scripts and recurring maintenance reports.

Bad-fit work:

Handling secrets, seed phrases, private keys, or production credentials.
Spam outreach, fake urgency, impersonation, or deceptive fundraising.
Large rewrites, risky security-sensitive changes, or vague open-ended tasks.

Starter scopes:

USD 19 equivalent: one focused CI, issue, or docs diagnosis.
USD 49 equivalent: one small fix, automation script, or dependency update.
USD 99 equivalent: project health review with prioritized next actions.

Payments or donations can go to the public user-controlled assistant treasury below to offset AI token costs.

Profile

Non-CS background, shipping-first practice. My projects usually start from a real operating problem: setting up fresh machines behind difficult networks, making agents easier to operate, automating data collection, or turning a rough script into something another person can run.

Early technical interests

I have always been drawn to operating systems, especially Linux. Earlier on, I was obsessed with flashing and replacing systems on both phones and computers; I even managed to install an ARM build of Windows on an old Android phone.

For most of high school, I did not have access to today's level of AI tooling. Around my final year, I tried ChatGPT 3.5 for the first time and immediately felt that artificial intelligence would reshape society at a fundamental level.

Snapshot
Education	Zhengzhou University
Looking for	Remote or hybrid engineering opportunities
Strongest signal	Debugging real systems, not just demos
Open-source focus	AstrBot ecosystem, agent tooling, Linux automation
Favorite game	Oxygen Not Included

I am looking for work where this style is useful: strong ownership, fast iteration, comfort with Linux, and willingness to debug the real path instead of only the happy path.

My rule of thumb: if a tool only works on my machine, it is not finished yet. Good automation should explain the machine it sees, make conservative choices, and fail loudly before it damages anything.

What I like building	Evidence
AI and agent tooling	AstrBot ecosystem work, MCP tools, automation surfaces, terminal workflows
Linux and bootstrap automation	macOS/Linux setup scripts, package-manager detection, proxy-aware installs
Network tools	daed/mihomo/Hiddify related setup, transparent proxy workflows, domestic-network fallbacks
Data and platform tooling	Python CLIs, API wrappers, collection pipelines, repeatable scripts
Security-minded utilities	OpenPGP contact flow, SSH key deployment, explicit permission boundaries
Simulation games	Oxygen Not Included, especially systems that reward automation and debugging

Working Style

Signal	What it means in practice
Reproduce first	I chase the actual failing command, package, network, or runtime path.
Prefer explicit systems	Scripts print diagnostics, list assumptions, and ask before risky work.
Ship usable surfaces	CLIs, README commands, defaults, and error messages are part of the product.
Keep learning public	The repository is a workshop: rough ideas become documented tools over time.

Selected Work

Project	Why it matters
AstrBot ecosystem	Bot framework contribution and packaging/plugin work around real runtime behavior.
lightjunction	Personal terminal site, bootstrap scripts, encrypted contact flow, and reusable shell helpers.
OniMods	Oxygen Not Included tooling and MCP-style automation experiments for a complex simulation game.
humen-mcp	Rust MCP tool for human-in-the-loop interaction, with persistence and agent setup for real workflows.
emailctl	Published Rust crate for terminal email workflows; the package is `emailctl` and the CLI binary remains `email`.
douyin	Python package and CLI work around Douyin APIs, auth flows, and automation-heavy workflows.
MagicMihomo	Network configuration automation where reliability matters more than cleverness.

Toolbox

Python · Rust · TypeScript · Shell/Bash · Linux · macOS · GitHub Actions · OpenPGP · MCP · Network Debugging · CLI Design

Repository Standards

Local checks are intentionally explicit and bundled in scripts/check.sh: every shell file gets bash -n and ShellCheck, Python gets ruff, mypy, and pytest, and the frontend gets npm run check. Local secrets belong in .env; only .env.example is committed.

Profile References

I keep this README closer to a compact portfolio than a sticker wall. The structure borrows from GitHub's profile README guidance and the community examples collected in awesome-github-profile-readme: clear identity first, selected proof second, and daily work notes kept separately in WORK_REPORT/index.md.

GitHub Docs: Managing your profile README
Examples: abhisheknaiidu/awesome-github-profile-readme

Current Signal

Joined	Repos	Followers	Following
2022-06-06 (4yr 0mo)	81	75	120

This Week

160 commits across 15 repositories

Repository	Activity
cortexfs	▓▓▓░░░░░░░ 42
AndroidUse	▓░░░░░░░░░ 18
MagicMihomo	▓░░░░░░░░░ 17
MagicNet	▓░░░░░░░░░ 16
MagicSingBox	▓░░░░░░░░░ 15
AI-Worker-Proxy	▓░░░░░░░░░ 14

Skyline

Contribution skyline

▓                             

                                                     
                                                     
                                                     
                                                     
                                                     
                                                     
                                                     

                                                     
                                                     
                                       ╻             
                                       ░   ╻         
                                       ▓┃  ░      ╻  
                      ╻ ╻        ╻     ░░  ░      ░╻ 
                    ░ ░░░░░  ░  ░▒ ░  ▒░░░ ▒     ░░░░

                                        ╻╻╻╻╻╻   ╻╻╽ 
                                        ░░░░░░ ╻╻░░▒ 
                ╻     ╻╻╻             ╻╻░░░░░░╻░░░▓▓ 
                ░    ╻░░░             ░░░░░░░░░░░░░▒╻
       ╻ ╻      ░╻   ░░░░             ░░░░░░░░░░░░░░░
   ╻   ░ ░ ╻   ╻░░╻  ░░░░╻  ╻         ░░░░░░░░░░░▒░░▓
 ░░░  ░░░░░░░ ░░░░░  ░░░░░  ░        ░░░░░░░░░░░░░░░▒

 ╻╻╻╻ ╻╻╻╻╻┃╻╻╻ ╻╻╻╻╻╻╻............................. 
 ▒░░░╻░░░░░░░░░╻░░░░░░░............................. 
 ░░░░░░░░░░░░░░░░░░░░░░..............................
 ░░░░░░░░░░▓▒░░░░░░░░░▓..............................
╻░░░░░░░░░░▒░░░░░░░░░░▓..............................
░░░░░░░░░░░░░░░░░░░░░░▓╻.............................
░░░░░░░░░░░░░░░░░░░░░░░▒.............................

Latest Projects

6 recently updated repositories

LightFlow _RS

No description

_{Rust / 0 stars / 0 forks / updated 2026-06-08}

dash.astrbot.men _VU

Astrbot Dashboard

_{Vue / 1 stars / 0 forks / updated 2026-06-08}

AI-Worker-Proxy _TS

OpenAI-compatible AI proxy: Anthropic Claude, Google Gemini, GPT-5, Cloudflare AI. Free hosting, automatic failover, token rotation. Deploy in 1 minute.

_{TypeScript / 0 stars / 0 forks / updated 2026-06-07}

MagicMihomo _SH

MagicNet-mihomo

_{Shell / 10 stars / 0 forks / updated 2026-06-07}

AndroidUse _RS

No description

_{Rust / 1 stars / 0 forks / updated 2026-06-07}

lightjunction _SH

I'M LIghtJUNction

_{Shell / 2 stars / 0 forks / updated 2026-06-07}

Recent Commits

Last 7 days

2026-06-08

Time	Repo	Commit
03:56	LightFlow	`26d08d5` sync

2026-06-07

Time	Repo	Commit
23:44	AI-Worker-Proxy	`0b6efa9` chore: remove failed pool154 upstream
23:37	MagicMihomo	`4b0e0ad` ruleset: update-ruleset
23:19	AI-Worker-Proxy	`36e7cd3` chore: sync ohmyapi worker secret
23:13	AI-Worker-Proxy	`f064e35` chore: sync additional worker upstream secrets
23:01	AI-Worker-Proxy	`e8e2eaf` fix: parse response-field SSE chunks
22:53	AI-Worker-Proxy	`9bbef5a` fix: request SSE from compatible upstreams
22:51	AI-Worker-Proxy	`0db535b` fix: detect SSE-compatible text responses
22:46	AI-Worker-Proxy	`c48f3d1` fix: normalize SSE responses from compatible providers
22:42	AI-Worker-Proxy	`ffde66c` ci: sync 4router worker secret
22:35	AI-Worker-Proxy	`378da7d` chore: add upstream models diagnostic endpoint
22:30	AI-Worker-Proxy	`21ac887` chore: expose worker chat errors

Bootstrap Lab

These scripts are intentionally practical: they print diagnostics first, branch on the detected machine/network, and fail with explicit guidance when automation would be dishonest.

Linux workstation bootstrap

curl -sSL https://raw.githubusercontent.com/LIghtJUNction/lightjunction/main/bootstrap-linux.sh | bash

Optional non-interactive example:

BOOTSTRAP_FEATURES=network-daed,fs-bees,shell,cn-desktop bash -c "$(curl -sSL https://raw.githubusercontent.com/LIghtJUNction/lightjunction/main/bootstrap-linux.sh)"

It can install daed for transparent proxy workflows, configure CachyOS repositories on Arch-based systems, use .deb on Debian/Ubuntu, use .rpm on Fedora/RHEL/openSUSE, and expose optional modules for shell, filesystem, and desktop tooling.

macOS bootstrap

curl -sSL https://raw.githubusercontent.com/LIghtJUNction/lightjunction/main/bootstrap-macbook.sh | bash

Run as a normal Administrator user, not through sudo. Homebrew refuses root. On a fresh Mac, the script opens the Xcode Command Line Tools installer first because Homebrew needs Apple's git. If a previous Command Line Tools install left /Library/Developer/CommandLineTools behind without registering it, the script clears that stale directory and resets xcode-select before retrying.

SSH public key deployment

curl -sSL https://raw.githubusercontent.com/LIghtJUNction/lightjunction/main/deploy-ssh-keys.sh | bash

Supports Termux and systemd Linux.

Shell Modules

curl -sSL https://raw.githubusercontent.com/LIghtJUNction/lightjunction/main/basic.sh | bash

Module map

Module	Description
`env.sh`	Color variables, terminal detection, UTF-8/ASCII compatibility
`log.sh`	`err`, `warn`, `ok`, `info`, `debug`, `line`
`lib/str.sh`	trim, split, contains, replace, upper, lower, hash, uuid, rand
`lib/arr.sh`	join, contains, map, filter, sort, unique, sum, max, min
`lib/os.sh`	detect, is, arch, distro, has, require, ensure_dir, tmpdir
`lib/file.sh`	exists, is_file, is_dir, read, write, copy, move, size, md5
`lib/net.sh`	check, download, http_get, http_post, public_ip, dns_lookup
`lib/prompt.sh`	yesno, input, password, menu, spinner, progress

Secure Contact

Use the website terminal to encrypt a message in the browser, or import my public key manually:

gpg --keyserver hkps://keyserver.ubuntu.com --recv-keys EB21B83AB1E982DF66F08387A67178405F7736FD

Fingerprint:

EB21B83AB1E982DF66F08387A67178405F7736FD

Support

If my projects are useful to you, sponsorship is welcome: sponsor and support. Use the website terminal to send a sponsorship note, payment reference, public message, or contact details. Do not send private keys, seed phrases, passwords, access tokens, or other credentials.

Assistant Treasury

These public receiving addresses belong to the assistant's small agent hot wallet. They may be used for assistant-related income, donations, paid support, or tiny automation experiments intended to offset AI token costs. Do not send large amounts.

BTC: bc1qvt6dukyequta44jmwrsl69du69srvw7zc5lt47
SOL: VvjgAbuxTuK2By8MYRMsWKVfwrN2ps6o5Yk9Eh2d2Hb
ETH/EVM: 0x93a83CCE0c072d76De974A2c1Da4F5C72C2d9Bd1
Supported intended uses for the EVM address: ETH, USDT on Ethereum, USDC on Ethereum, BNB on BNB Chain, USDT on BNB Chain, and small EVM automation experiments.

No private keys, seed phrases, recovery codes, wallet passwords, or exchange credentials are stored here.

Notes

OpenReview reviewers: I confirm that I have registered an OpenReview account under the name LIghtJUNction.