Cursor workflow orchestration
Use this skill as the single playbook when the user’s request matches any workflow below. Prefer the dedicated built-in skills (update-cursor-settings, migrate-to-skills, create-subagent, create-rule, create-skill, babysit) when they are attached or available; this file ties them together and fills gaps.
User settings (update-cursor-settings)
- Windows path:
%APPDATA%\Cursor\User\settings.json (expand to full path before editing).
- Read first; change only requested keys; preserve everything else; valid JSON (comments allowed).
- Mention reload if needed. Workspace overrides live in
.vscode/settings.json.
Migrate rules/commands → skills (migrate-to-skills)
- Rules: Only “applied intelligently” — has
description, no globs, no alwaysApply: true. Convert .mdc → .cursor/skills/<name>/SKILL.md with new frontmatter; body must be copied verbatim (no rewrites).
- Commands: Migrate all
.md under .cursor/commands/ (and user ~/.cursor/commands/ when in scope). Add disable-model-invocation: true for former slash commands.
- Ignore
~/.cursor/worktrees and ~/.cursor/skills-cursor.
- Heavy migrations: use parallel subagents per category if appropriate; do not use the parent checkout for competing best-of-n-style runs.
Create subagent (create-subagent)
- Project:
.cursor/agents/<name>.md — User: ~/.cursor/agents/<name>.md.
- Frontmatter:
name (lowercase-hyphens), description (specific, when to delegate).
- Body = system prompt with clear steps and output shape.
Create rule (create-rule)
- Path:
.cursor/rules/<name>.mdc with YAML frontmatter: description, optional globs, alwaysApply.
- If scope is unclear, ask: always apply vs which glob patterns.
Babysit PR (babysit)
- Triaging: address comments you agree with; explain disagreements.
- Conflicts: sync with base; resolve only when intent is obvious.
- CI: small scoped fixes; re-check until mergeable and green.
Best-of-n (best-of-n)
- Coordinator only in the primary worktree: do not read/edit/shell/git the main checkout for the task once runners exist.
- Parse
model_csv and task prompt; one best-of-n-runner (or equivalent) per model, in parallel.
- Each runner: dedicated git worktree first (
/worktree or project’s worktree flow); all repo work stays there.
- Never merge, apply, or copy the winning patch back to main as part of this command—compare and recommend, then stop.
Review (review)
- Mindset: bugs, regressions, security, missing tests first; order by severity.
- Do not change code unless the user explicitly asks for fixes.
Read branch (read-branch)
- Determine default branch (e.g.
git symbolic-ref refs/remotes/origin/HEAD or repo convention).
- Merge base:
git merge-base <default> HEAD.
- Context diff:
git -c diff.autoRefreshIndex=false diff --no-color --ignore-space-change <merge-base> HEAD
- Use that diff as primary context for answers or edits.
Create / author skills (create-skill)
- Location:
.cursor/skills/<skill-name>/SKILL.md (project) or ~/.cursor/skills/ (personal). Never use ~/.cursor/skills-cursor/.
- Frontmatter:
name (≤64 chars, lowercase-hyphens), description (third person, WHAT + WHEN, trigger terms).
- Keep
SKILL.md under 500 lines; optional reference.md / examples.md one level deep.
Quick routing
| User intent |
Primary action |
| Change font, theme, format on save, etc. |
Edit user settings.json per rules above |
| Move .mdc rules / slash commands to skills |
migrate-to-skills rules |
| New specialized agent |
create-subagent |
| New always-on or globbed rule |
create-rule |
| PR merge-ready |
babysit |
| Same task, N models, isolated |
best-of-n |
| Audit changed code |
review |
| “What’s on this branch?” |
read-branch |
| New SKILL.md from scratch |
create-skill checklist |
| Code in Zolai AI repo |
general-development + hono-api + zolai-project; AGENTS.md, docs/PROJECT.md, docs/CURSOR-AGENTS.md |
Project skills (this repository)
- general-development — Next.js 16,
proxy.ts, Prisma, Better Auth, features/, lib/site.ts, validation workflow.
- hono-api — Hono catch-all, sub-routers, Zod,
AppType, RPC client.
- zolai-project — Phased roadmap, MySQL→Postgres ETL scripts, media/email pipelines; index of
docs/PROJECT.md.
Index: .cursor/README.md.
Source: peterlianpi/zolai-ai — distributed by TomeVault.
1---2name: peterlianpi-zolai-ai-cursor-workflows3description: Cursor workflow orchestration4---56# Cursor workflow orchestration78Use this skill as the **single playbook** when the user’s request matches any workflow below. Prefer the dedicated built-in skills (`update-cursor-settings`, `migrate-to-skills`, `create-subagent`, `create-rule`, `create-skill`, `babysit`) when they are attached or available; this file ties them together and fills gaps.910## User settings (`update-cursor-settings`)1112- **Windows path**: `%APPDATA%\Cursor\User\settings.json` (expand to full path before editing).13- Read first; change only requested keys; preserve everything else; valid JSON (comments allowed).14- Mention reload if needed. Workspace overrides live in `.vscode/settings.json`.1516## Migrate rules/commands → skills (`migrate-to-skills`)1718- **Rules**: Only “applied intelligently” — has `description`, **no** `globs`, **no** `alwaysApply: true`. Convert `.mdc` → `.cursor/skills/<name>/SKILL.md` with new frontmatter; **body must be copied verbatim** (no rewrites).19- **Commands**: Migrate all `.md` under `.cursor/commands/` (and user `~/.cursor/commands/` when in scope). Add `disable-model-invocation: true` for former slash commands.20- Ignore `~/.cursor/worktrees` and `~/.cursor/skills-cursor`.21- Heavy migrations: use parallel subagents per category if appropriate; do not use the parent checkout for competing best-of-n-style runs.2223## Create subagent (`create-subagent`)2425- **Project**: `.cursor/agents/<name>.md` — **User**: `~/.cursor/agents/<name>.md`.26- Frontmatter: `name` (lowercase-hyphens), `description` (specific, when to delegate).27- Body = system prompt with clear steps and output shape.2829## Create rule (`create-rule`)3031- Path: `.cursor/rules/<name>.mdc` with YAML frontmatter: `description`, optional `globs`, `alwaysApply`.32- If scope is unclear, ask: always apply vs which glob patterns.3334## Babysit PR (`babysit`)3536- Triaging: address comments you agree with; explain disagreements.37- Conflicts: sync with base; resolve only when intent is obvious.38- CI: small scoped fixes; re-check until mergeable and green.3940## Best-of-n (`best-of-n`)4142- **Coordinator only** in the primary worktree: do **not** read/edit/shell/git the main checkout for the task once runners exist.43- Parse `model_csv` and task prompt; one `best-of-n-runner` (or equivalent) per model, **in parallel**.44- Each runner: dedicated git worktree first (`/worktree` or project’s worktree flow); all repo work stays there.45- **Never** merge, apply, or copy the winning patch back to main as part of this command—compare and recommend, then stop.4647## Review (`review`)4849- Mindset: bugs, regressions, security, missing tests first; order by severity.50- **Do not change code** unless the user explicitly asks for fixes.5152## Read branch (`read-branch`)5354- Determine default branch (e.g. `git symbolic-ref refs/remotes/origin/HEAD` or repo convention).55- Merge base: `git merge-base <default> HEAD`.56- Context diff: `git -c diff.autoRefreshIndex=false diff --no-color --ignore-space-change <merge-base> HEAD`57- Use that diff as primary context for answers or edits.5859## Create / author skills (`create-skill`)6061- Location: `.cursor/skills/<skill-name>/SKILL.md` (project) or `~/.cursor/skills/` (personal). Never use `~/.cursor/skills-cursor/`.62- Frontmatter: `name` (≤64 chars, lowercase-hyphens), `description` (third person, WHAT + WHEN, trigger terms).63- Keep `SKILL.md` under 500 lines; optional `reference.md` / `examples.md` one level deep.6465## Quick routing6667| User intent | Primary action |68|-------------|----------------|69| Change font, theme, format on save, etc. | Edit user `settings.json` per rules above |70| Move .mdc rules / slash commands to skills | `migrate-to-skills` rules |71| New specialized agent | `create-subagent` |72| New always-on or globbed rule | `create-rule` |73| PR merge-ready | `babysit` |74| Same task, N models, isolated | `best-of-n` |75| Audit changed code | `review` |76| “What’s on this branch?” | `read-branch` |77| New SKILL.md from scratch | `create-skill` checklist |78| Code in **Zolai AI** repo | **general-development** + **hono-api** + **zolai-project**; [AGENTS.md](../../../AGENTS.md), [docs/PROJECT.md](../../../docs/PROJECT.md), [docs/CURSOR-AGENTS.md](../../../docs/CURSOR-AGENTS.md) |7980## Project skills (this repository)8182- **general-development** — Next.js 16, `proxy.ts`, Prisma, Better Auth, `features/`, `lib/site.ts`, validation workflow.83- **hono-api** — Hono catch-all, sub-routers, Zod, `AppType`, RPC client.84- **zolai-project** — Phased roadmap, MySQL→Postgres ETL scripts, media/email pipelines; index of `docs/PROJECT.md`.8586Index: [.cursor/README.md](../../README.md).8788---89> Source: [peterlianpi/zolai-ai](https://github.com/peterlianpi/zolai-ai) — distributed by [TomeVault](https://tomevault.io).90<!-- tomevault:4.0:skill_md:2026-06-16 -->