Tips: Linux, Networking, Troubleshooting & AI 🐧🐧🐧

A practical reference for the command line. Commands tested on Arch Linux / systemd systems; most apply to any Linux distro.

Linux Tips (1–40)

# Tip Command / Notes
1 Search command history interactively Ctrl+R — type to filter, press again to cycle matches
2 Repeat the last command !! — useful as sudo !! when you forgot sudo
3 Reuse the last argument of the previous command !$ e.g. mkdir /long/path && cd !$
4 Jump to the previous directory cd -
5 Directory stack for multi-directory navigation pushd /path / popd / dirs -v
6 Search inside a man page Press /pattern then n / N to cycle results
7 Search backwards in less ?pattern inside less
8 Recursive text search with line numbers grep -rn "pattern" /path — add -l for filenames only
9 Find files modified in the last 7 days find . -name "*.log" -mtime -7
10 Pass command output as arguments to another command find . -name "*.bak" | xargs rm
11 Write output to a file and still see it on screen command | tee output.txt
12 Extract a specific column from text output awk '{print $2}' file — change $2 for the column you need
13 Global find-and-replace inside a file sed -i 's/old/new/g' file
14 Count occurrences, most frequent first sort | uniq -c | sort -rn
15 Disk usage of every item in the current directory du -sh *
16 Human-readable disk free space df -h
17 Find what is using a specific port lsof -i :8080 or ss -tulnp | grep 8080
18 Find a running process by name pgrep -a nginx or ps aux | grep nginx
19 Kill a process gracefully, then forcefully if needed kill PID (SIGTERM first) → kill -9 PID (SIGKILL)
20 Keep a session alive after logout tmux new -s name / tmux attach -t name
21 Create reusable command shortcuts alias ll='ls -lh --color=auto' — add to ~/.zshrc
22 Search history for a specific command type history | grep ssh
23 Suspend a foreground process and background it Ctrl+Z then bg
24 List background jobs and bring one to foreground jobs → fg %1
25 Make a script executable chmod +x script.sh
26 Change file owner and group chown user:group file
27 Create a symbolic link ln -s /path/to/target linkname
28 Compress a directory into a tarball tar -czf archive.tar.gz dir/
29 Extract a tar archive tar -xzf archive.tar.gz — add -C /dest for a target directory
30 Sync files efficiently with progress rsync -avz --progress src/ dest/
31 Generate a modern SSH key pair ssh-keygen -t ed25519 -C "your@email"
32 Copy your public key to a remote host ssh-copy-id user@host
33 Follow system logs in real time journalctl -f — add -u nginx to filter by unit
34 Manage systemd services systemctl status|start|stop|restart|enable|disable service
35 View recent kernel messages with timestamps sudo dmesg -T | tail -n 50
36 Trace system calls of a running process strace -p PID
37 Show shared library dependencies of a binary ldd /usr/bin/binary
38 Locate where a command lives which cmd / whereis cmd / type cmd
39 Abort on any error in bash scripts set -euo pipefail at the top of every script
40 Run a command that survives terminal close nohup command & disown

↑ Back to top

Networking Tips (41–70)

# Tip Command / Notes
41 Show all network interfaces and addresses ip addr (modern replacement for ifconfig)
42 View the routing table ip route
43 List open ports and their processes ss -tulnp (modern replacement for netstat -tulnp)
44 Detect service versions on open ports nmap -sV host
45 Scan all 65535 ports nmap -p- host
46 Limit ping to a fixed count ping -c 4 host
47 Trace the route to a host traceroute host or tracepath host (no root needed)
48 Fetch only HTTP response headers curl -I https://example.com
49 Debug a curl request in full detail curl -v https://example.com
50 Resume an interrupted download wget -c URL
51 Full DNS record lookup dig domain ANY
52 Quick, clean DNS answer dig +short domain
53 Query a specific DNS server dig @8.8.8.8 domain
54 Domain registration and owner info whois domain
55 List current firewall rules iptables -L -n -v or nft list ruleset (nftables)
56 Capture live traffic on an interface tcpdump -i eth0 -n port 80
57 Quick port scan with netcat nc -zv host 20-25
58 Simple TCP listener for connectivity testing nc -l 4444 (server) / nc host 4444 (client)
59 Forward a local port through SSH ssh -L 8080:localhost:80 user@host
60 Create a SOCKS proxy via SSH ssh -D 1080 user@host — point browser to localhost:1080
61 Real-time traceroute with packet loss stats mtr host
62 View the ARP cache arp -n or ip neigh
63 Get all local IP addresses hostname -I
64 Get your public IP from the terminal curl -s ifconfig.me
65 Override DNS for a host locally Add 1.2.3.4 hostname to /etc/hosts
66 Check or change the DNS resolver /etc/resolv.conf — on systemd: resolvectl status
67 Manage connections via NetworkManager CLI nmcli con show / nmcli dev wifi connect SSID password PASS
68 Scan for nearby WiFi networks nmcli dev wifi list or iwlist wlan0 scan
69 Check NIC speed and duplex ethtool eth0
70 Save a packet capture for Wireshark tcpdump -i eth0 -w capture.pcap

↑ Back to top

Top 10 Nmap Commands

# Use Case Command
1 Quick ping scan — discover live hosts on a subnet nmap -sn 192.168.1.0/24
2 Fast scan of the 1000 most common ports nmap -F host
3 Scan all 65535 ports nmap -p- host
4 Detect service versions on open ports nmap -sV host
5 OS detection (requires root) sudo nmap -O host
6 Aggressive scan — OS, version, scripts, traceroute sudo nmap -A host
7 SYN stealth scan (half-open, requires root) sudo nmap -sS host
8 UDP scan for common services (DNS, SNMP, etc.) sudo nmap -sU --top-ports 20 host
9 Run default NSE scripts against a host nmap -sC host
10 Save results in all formats (normal, XML, grepable) nmap -oA scan_output host

↑ Back to top

Top 10 Wireshark Display Filters

# Use Case Filter
1 Show only HTTP traffic http
2 Show only HTTPS / TLS traffic tls
3 Filter by source or destination IP ip.addr == 192.168.1.10
4 Filter by source IP only ip.src == 192.168.1.10
5 Filter traffic on a specific port tcp.port == 443 or udp.port == 53
6 Show only DNS queries and responses dns
7 Show TCP connection resets (useful for troubleshooting drops) tcp.flags.reset == 1
8 Show TCP SYN packets only (spot connection attempts) tcp.flags.syn == 1 && tcp.flags.ack == 0
9 Filter traffic between two specific hosts ip.addr == 10.0.0.1 && ip.addr == 10.0.0.2
10 Search packet contents for a string (e.g. credentials in cleartext) frame contains "password"

↑ Back to top

Troubleshooting Tips (71–100)

# Tip Command / Notes
71 Always start with the logs journalctl -xe or /var/log/syslog / /var/log/messages
72 Find kernel errors and warnings dmesg | grep -i "error\|fail\|warn"
73 Check available memory free -h
74 Real-time CPU, memory, and I/O stats vmstat 1 — updates every second
75 Disk I/O utilization per device iostat -x 1
76 Interactive process monitor htop (install separately) or top
77 See which process is hammering the disk iotop (requires root)
78 Check system load averages uptime — 1/5/15-min averages; above CPU count = overloaded
79 See recent successful logins last
80 See failed login attempts lastb (requires root)
81 Check auth log for brute force activity grep "Failed password" /var/log/auth.log
82 Get a summary of network statistics ss -s
83 List all block devices and mount points lsblk
84 Get UUIDs of block devices for /etc/fstab blkid
85 Check and repair a filesystem fsck /dev/sdXN — only on unmounted partitions
86 Check drive health via SMART smartctl -a /dev/sda — install smartmontools
87 See exactly what a command does at the OS level strace -e trace=file command
88 Trace library calls ltrace command
89 Check if resource limits are causing failures ulimit -a — raise open files with ulimit -n 65535
90 See which process owns each network socket ss -tp
91 Find what is preventing an unmount lsof +D /mnt/drive or fuser -m /mnt/drive
92 View all environment variables env or printenv VAR for a single variable
93 Fix a broken or garbled terminal reset — or stty sane if reset is also unresponsive
94 Determine what type of file something is file /path/to/file
95 Check the HTTP status code of a URL curl -o /dev/null -sw "%{http_code}\n" URL
96 Test if a remote port is reachable nc -zv host port
97 Check SSL certificate expiry date echo | openssl s_client -connect host:443 2>/dev/null | openssl x509 -noout -dates
98 Find large files consuming disk space find / -type f -size +100M -exec ls -lh {} \; 2>/dev/null | sort -k5 -rh
99 Verify a file's integrity with a checksum sha256sum file then compare against the published hash
100 Always capture verbose output before escalating an issue command -v 2>&1 | tee debug.log — attach the log when asking for help

↑ Back to top

AI Tips (Top 30)

# Tip Notes
1 Be specific in your prompts Vague prompts get vague answers. Include context, constraints, and the exact output format you want.
2 Use a system prompt to set the model's role Tell the model who it is before asking questions: "You are a senior Linux sysadmin..." shapes every response that follows.
3 Ask the model to think step by step Adding "think step by step" or "reason through this before answering" significantly improves accuracy on complex tasks.
4 Give examples of what you want (few-shot prompting) Show the model 2–3 input/output examples before your real request. Dramatically improves formatting and style consistency.
5 Ask for multiple options, then choose "Give me 3 different approaches to X" — comparing options leads to better decisions than taking the first answer.
6 Tell the model what NOT to do Negative constraints work: "Do not use jargon", "Do not include code examples", "Do not summarize at the end."
7 Break large tasks into smaller prompts Chaining focused prompts outperforms one giant prompt. Do one thing well, then pass the output to the next step.
8 Ask the model to critique its own output Follow up with "What are the weaknesses of that answer?" or "What did you miss?" — models often self-correct well.
9 Use temperature 0 for factual or code tasks Low temperature = more deterministic output. Use higher values (0.7–1.0) for creative writing, brainstorming.
10 Always verify code before running it AI-generated code can be plausible-looking but wrong. Read it, test it in a safe environment, never run blindly as root.
11 Use RAG to ground the model in your own data Retrieval-Augmented Generation lets you feed relevant documents into the prompt at query time — reduces hallucination on domain-specific questions.
12 Keep context windows clean Old, irrelevant messages dilute attention. Start a new conversation when switching topics rather than continuing a stale thread.
13 Use structured output (JSON/XML) for programmatic use Ask for JSON output and specify the schema: "Respond only with JSON matching this structure: {...}". Most models comply reliably.
14 Anchor the model with facts it cannot hallucinate around Paste in the actual documentation, log output, or code you want analyzed. Don't ask the model to recall things from training data alone.
15 Use a local model for sensitive data Tools like Ollama let you run Llama, Mistral, and others locally — nothing leaves your machine. Essential for private data or air-gapped environments.
16 Embed with the same model you query When building vector search, generate embeddings with the same model (or family) you use for generation — mixing models degrades retrieval quality.
17 Use AI as a first-pass code reviewer Paste a diff or function and ask "What bugs or security issues do you see?" Catches obvious problems before a human review.
18 Prompt for explanations, not just answers "Explain why" forces the model to surface its reasoning. If the explanation doesn't make sense, the answer probably doesn't either.
19 Use AI to generate test cases "Generate edge case inputs for this function" is one of the highest-value AI tasks in a development workflow.
20 Iteratively refine — don't rewrite the whole prompt If output is 80% right, say "Keep everything but change X to Y." Incremental corrections preserve what's working.
21 Use AI to translate between formats CSV → JSON, YAML → TOML, regex → plain English explanation, SQL → pandas — format conversion is a reliable strength.
22 Fine-tuning is usually the last resort Better prompting or RAG solves most problems cheaper. Fine-tune only when you need style/format consistency at scale that prompting cannot provide.
23 Check token limits before sending large inputs Content silently truncated at the context limit causes subtle errors. Know your model's limit and chunk inputs if needed.
24 Use AI to summarize long documents before asking questions For documents near the context limit, ask for a structured summary first, then ask questions against the summary.
25 Build evaluation sets before you build pipelines Define what "good output" looks like with 10–20 examples before automating. Without eval criteria, you cannot tell if changes help or hurt.
26 Log every prompt and response in production AI output is non-deterministic. Without logs you cannot reproduce failures, audit decisions, or improve prompts over time.
27 Use AI agents for multi-step tasks, not single queries Frameworks like CrewAI or AutoGen let agents plan, use tools, and loop — appropriate when the answer requires multiple actions, not just one generation.
28 Never trust AI output on security-critical decisions AI can produce convincing but wrong security advice. Always cross-reference CVEs, advisories, and documentation from authoritative sources.
29 Use AI to accelerate learning, not replace it Ask AI to explain concepts, generate practice problems, and review your work — but do the thinking yourself. Understanding the output is what builds skill.
30 Know when not to use AI Simple lookups (man pages, docs), real-time data, and tasks requiring guaranteed accuracy are often faster and safer without AI in the loop.

↑ Back to top