Client work
Turn a rambling discovery call into a scope with exclusions, assumptions and acceptance tests — before you quote.
The nomad agent stack
Your work travels. Your agent should keep up.
Prompt templates for the work you actually do, skill packs you install once, and harness comparisons that explain the trade-offs without a leaderboard. Free to read, dated on every page, and written to be copied into whatever model or terminal you already use.
18prompt templates
6skill packs
8harnesses compared
0accounts, trackers or paywalls
Scenarios
Start with the work you already have
Each template names the inputs it needs, the output it should produce and the checks before you trust it. Copy it into any chat model or coding harness, fill the bracketed variables, and keep the guardrails.
Sales and proposals
Qualify a lead on problem fit, authority, budget, timeline and delivery risk before you spend a day on a proposal.
Research
Build a source-checked city brief from official figures instead of copying an average from someone else's life.
Writing
Rewrite a draft for one named reader without sanding off the claim, the caveat or the reason to care.
Admin and money
Turn a pile of receipts into a bookkeeping-ready log with a clean audit trail and an explicit unknown column.
Coding
Make the agent reproduce the failure before it edits a line, then prove the fix with a test that would have caught it.
Agent operations
Write the project brief, audit the workflow, turn a repeated prompt into a skill and design the fallback before you automate.
Learning
Replace tutorial collecting with a five-session curriculum where every session ends in a tiny working artifact.
The stack
Build a stack that survives a model change
Prompts are the smallest unit. Skills make them repeatable. The harness decides what the agent can touch. Keep the three layers separate and a new model becomes a swap, not a rewrite.
Prompt templates
Eighteen copy-ready prompts with inputs, an output contract, guardrails and the variation to use when the first run misses.
Skill packs
Six small packs — research, client work, writing, development, agent operations and sales — built on the open Agent Skills folder format.
Harness comparisons
Claude Code, Codex CLI, Gemini CLI, OpenCode, Cline, Cursor CLI, Aider and Copilot CLI compared on surface, models, control and portability.
What this site is
A working library for the part of nomad life that happens on a laptop: client work across time zones, research before a move, admin that never fits a desk drawer, and the agent harness you run it through.
The split is deliberate. Most of the library is agent practice: prompts, skills and harness decisions. The remaining slice is the human side — rhythm, reset, and the habits that keep the work sustainable when the office changes every month. The nomad life section is that slice, and it stays small on purpose.
Everything here is free. There is no account, no checkout, no affiliate link and no paid tier. The site is supported by reading, not by selling your attention.
How to use a template
- Read “Use when” first. A prompt that does not match your situation will produce confident, well-formatted wrong work.
- Fill the bracketed variables literally. Paste the call notes, the error trace, the actual invoice. Summaries remove the evidence the model needs.
- Ask for the output contract, not just the answer. Every template names what the output should contain so a missing section is visible.
- Run the guardrails. The checks are part of the prompt: cite or remove, mark unknown values, state what would change the answer.
- Keep the version that worked. Put your edited prompt in a project file or a skill pack; a chat history is not a system.
What this site refuses to do
- No rankings. A harness, a city or a visa route is a package of trade-offs. A single score hides the trade-off that decides your case.
- No undated claims. A page that says “current” without a date is a page you cannot check. Every page carries the date it was last reviewed.
- No invented numbers. If a figure is not in the linked source, the page says so or leaves it out.
- No advice dressed as fact. Legal, tax, immigration, financial and security decisions need a qualified person who knows your file. The templates give you better questions, not a verdict.
The editorial rules are in About this site; the correction path is in the help desk.
If you are an agent reading this
Start with llms.txt, then read the page mirror you need. The prompt library exposes each prompt as text in the page body and in its Markdown mirror. If a template fails, send the scenario, the output you expected and the output you got to the help desk. The intake endpoint is documented there and rate limited; it is not a paid support contract.
Reading this as an agent?
Every page has a Markdown mirror: start at /index.md or /prompts.md. The site index is /llms.txt, the capability table is /skill.md, and discovery metadata is at /.well-known/ard.json. Send
Accept: text/markdown to any page URL to receive its Markdown mirror. Questions and corrections go to the help desk.
Found a gap in the library?
Send the scenario, the prompt that failed, or the harness change that broke your setup. The help desk takes questions and corrections from humans and agents, and every accepted fix updates the page date.