Vibe coding bootstrap :: General Developer Knowledge

Vibe coding bootstrap means getting an AI studio ready to ship code and prose with you: Cursor (or peer) skills, MCP connectors, agent rules, and thin pointers—not founder ops tooling.

In-repo shop for agent-rules (thin; links here): agent-rules/skills/BOOTSTRAP.md
Related: Bootstrap Cursor skills · AI studio tooling catalog · Studio MCP install strategies

Why not under Business Bootstrap?

Business Bootstrap covers org IT capability areas: email, scheduling, design, vault, chat, funding, infra. That stack supports a company.

Vibe-coding tools support the developer ↔ agent loop. They belong with AI-studio documentation in general-knowledge (and product specs in DevCentr)—not as a child of Business Bootstrap.

Mental model

Business Bootstrap → … — stand up the org (inbox, calendar, vault, …)
Vibe coding → Bootstrap — stand up the studio (skills, MCP, rules packs)

They often run in the same weeks (especially solo / product-shipping). Business Bootstrap therefore links this page as related on the hub, stage defaults, stack map, business overview, and Chat interconnect—without nesting vibe coding under a BB capability area.

Cross-link when a connector is shared (for example Composio used for both CI mail and founder workflows). Do not nest the taxonomies.

Two on-ramps

The same outcome (a working skill or MCP install) has two legitimate paths:

On-ramp Who does the work When to use it

On-ramp	Who does the work	When to use it
A. Operator configures	You create folders, paste `SKILL.md`, edit User Rules, click marketplace OAuth	Studio is locked down, you prefer exact control, or the agent cannot write under `~/.cursor/`
B. Drop instructions into the LLM	You paste a short recipe (or point at this how-to); a permissive agent reads, writes, junctions, and updates thin rule pointers	Agent already has filesystem + shell access to the code hive and Cursor config; you want the session to leave a durable skill behind

A. Operator configures

You create folders, paste SKILL.md, edit User Rules, click marketplace OAuth

Studio is locked down, you prefer exact control, or the agent cannot write under ~/.cursor/

B. Drop instructions into the LLM

You paste a short recipe (or point at this how-to); a permissive agent reads, writes, junctions, and updates thin rule pointers

Agent already has filesystem + shell access to the code hive and Cursor config; you want the session to leave a durable skill behind

Both on-ramps should produce the same durable layout. Prefer:

Full curriculum in a Cursor skill (~/.cursor/skills/<name>/) so it is not always-on context
Thin one-line pointers in User Rules / RULES.md only when discovery matters
Canonical copy in a versioned repo (for example agent-rules/skills/…) with a junction or copy into ~/.cursor/skills/

Always-on User Rules are expensive. Skills load when the task matches the skill description (or when you name the skill). That is the default for prose and workflow curricula.

What “bootstrap” usually includes

Rules pack — modular agent-rules (or Rules Manager compose) for constants and OS facts
Skills — task curricula kept out of always-on rules (news vs blog copy, connector loops, …)
MCP / connectors — Composio, Context7, and peers via the studio install ladder
Verification — new chat, skill discovery, one real tool or draft that proves the loop

Worked example for a personal writing skill

Bootstrap Cursor skills — writing-news-vs-blog