Foreman
Every good build starts as a drawing.
Foreman makes your agent do the drawing first: the interview, the decisions, the locked system, the brief. Then the code, then the domain, then indexed.
It answers with questions, not code.
> help me build my portfolio Before any code, I need two things from you: what you are building, and what you already have for it. Copy, CV, screenshots, a logo, a domain, anything. Then we make the decisions that decide whether this looks like yours or like every other generated site.
Without a playbook, that same message returns a finished page built on assumptions you never made.
Same agent, same model, different session.
| Without Foreman | With Foreman | |
|---|---|---|
| First move | Generates a page | Asks what the site is for and who reads it |
| Copy | Written by the model | Extracted from you, with numbers |
| Design | Whatever the model defaults to | Locked tokens you chose from three variants |
| Scope | Everything it can think of | Explicit non-goals, written before the build |
| Done | It renders on your laptop | 375px, keyboard, Lighthouse, incognito, real 404 |
| Live | "Deployed" | DNS, SSL, sitemap, Open Graph, submitted and indexed |
Seven phases. Each has a gate it will not pass.
Your agent runs this. You do not read it, and it does not narrate it at you.
It knows the subject and your asset inventory.
One job, one action, a register, non-goals, and draft copy for every section, in writing.
Exact hex values, exact font names, a type scale, and one named signature element.
A frozen brief you have read and approved.
It builds with zero errors and you have seen it at 375px.
Evidence, not assurances. Then the domain resolves over HTTPS in incognito.
The sitemap is submitted and the shared link renders its card.
Markdown and templates. Nothing executes.
No scripts, no hooks, no MCP servers, no network calls. You can read every line before you trust it, which is the only sane way to install something that talks to your agent.
References
Loaded only when the phase arrives: the content interview, design direction, stack choice, the brief, performance and accessibility, metadata and 404, bilingual and RTL, verify and ship, and a full worked example.
Fillable assets
A brand harness with three live variants, a head block with Open Graph and JSON-LD, a custom 404, robots.txt, sitemap.xml, and llms.txt.
Failure modes
The places builds actually die, watched for throughout: delegated decisions, missing non-goals, big-bang prompting, trusting the desktop render, machine translation, DNS edits made blind.
One folder of markdown. It goes anywhere an agent reads files.
Claude Code
/plugin marketplace add Turki-Sh/Foreman /plugin install foreman@alshuaibi
Cursor · Windsurf · Codex · Copilot · Gemini CLI · any coding agent
git clone https://github.com/Turki-Sh/Foreman.git cp -r Foreman/plugins/foreman/skills/foreman ~/.agents/skills/
Per-tool paths are in AGENTS.md.
Claude apps, or any assistant with no skills support
Upload foreman.skill, or paste SKILL.md into the conversation and say "run this on me."
Then say
help me build my portfolio
It should answer with two questions, not with code. If it starts writing HTML, it did not load.