agentsclimarketplace
Sign in

Project vibe master workflow

Skill wchen02/cursor-agent-learning/.cursor/skills/project-vibe-master-workflow

End-to-end project-level vibe workflow with split on-disk state under .cursor/project-vibe-master/ for resume; state template and optional generated data live in this skill (initial-state/, data/). Combines rules, shell/scripts, MCP, hooks, subagents, HITL, and review/deploy skills.From its SKILL.md

Install
npx -y skills add wchen02/cursor-agent-learning --skill project-vibe-master-workflow

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

7.5 KB, ~1.8k tokens by cl100k_base, as published. Nobody here has run it

Project vibe master workflow (full stack)

Orchestration skill. This workflow does not replace the focused skills (/agile-project-lifecycle-mcp, /release-with-approval, etc.); it chains them conceptually so one session can exercise rules + scripts + MCP + hooks + subagents + HITL + review/deploy skills.

Apply @general-conventions, @mcp-patterns (when using MCP), and @human-in-the-loop at gates. For file-type rules, add @typescript-patterns / @api-conventions as the stack requires.

On-disk state (split files + resume) — do this first, every run

  1. Default directory: .cursor/project-vibe-master/ (folder under project root, next to other .cursor/ config). If the user names another directory, use it consistently.
  2. Start of session — If the folder does not exist, create it by copying the template tree initial-state/ (bundled next to this skill) to .cursor/project-vibe-master/ (entire directory). Set meta.md (project one-liner, last updated) and initial checkboxes in progress.md.
  3. Read order — project-vibe-master/README.md (layout), then meta.md, progress.md, gates.md, slices/README.md, only the active slice file(s) listed there, backlog.md, top of sessions.md.
  4. Resume logic — From progress.md (snapshot + phase checklist), then gates.md (do not re-ask a gate already approved unless the user asks to revisit), then slices/ (first slice with an unchecked critical step = current work; current slice file is named in slices/README.md and progress.md).
  5. After any meaningful step — Update the smallest file that fits (see README “Write rules” in that folder): e.g. slice work → only that slices/NN-slug.md; gate reply → gates.md + progress.md; session wrap → prepend sessions.md; always bump Last updated in meta.md when you change state.
  6. At each HITL stop — Set Blocked on in progress.md before stopping.
  7. User visibility — When resuming, summarize Completed / In progress / Next up from progress.md + active slice file so the user can correct you.
  8. Generated artifacts — Put exports, scratch captures, or MCP dumps from practicing this workflow in data/ (see data/README.md); keep initial-state/ unchanged as the clean template.

Legacy: If you find an old single file .cursor/project-vibe-master-state.md, migrate its sections into the split layout (or ask the user); prefer the directory layout for new work.

When to use

  • Greenfield or large change where you want requirements → iterative build → tests → human feedback → security/review → deploy checklist → release gate in one narrative.
  • Learning: run through every technique this repo demonstrates in order without editing the underlying skills.
  • Stop/start: New chat — rerun the skill; it picks up from .cursor/project-vibe-master/.

Phase 0 — Pre-flight (orientation, no code yet)

  1. Rules — Load conventions above; mention where MCP/hooks apply.
  2. Shell (safe demos) — From repo root, optionally run ./examples/scripts/demo-info.sh and ./examples/scripts/count-docs-md.sh (read-only; see tool-script-calling).
  3. MCP (optional) — If servers are configured: check tool schema first; e.g. project-docs list_example_scripts to list script docs without executing on the server. For heavy doc fetch, optionally delegate mcp-docs-helper subagent.
  4. Hooks (awareness) — If the user enabled them: beforeShellExecution / afterShellExecution (audit/block shell) and beforeMCPExecution / afterMCPExecution (audit MCP); see .cursor/hooks/README.md — do not modify hook files in this skill.
  5. State — Mark Phase 0 in progress.md; update progress.md snapshot; meta.md timestamp.

Phase 1 — Discovery, requirements, Gate 1 (HITL)

  1. Research — explore subagent (or main agent): map codebase and patterns.
  2. External truth — MCP for live docs/APIs where relevant; follow @mcp-patterns.
  3. Requirements — User outcomes, acceptance criteria, out of scope, constraints.
  4. Optional security baseline — /security-audit (quick pass) or security-auditor subagent on sensitive areas (readonly).
  5. Gate 1 — Requirements — Explicit handoff: Approve requirements / Refine / Abort. Stop until the human replies.
  6. State — Optional requirements pointer in sessions.md or slice notes; before stop, Blocked on in progress.md; after human replies, gates.md + progress.md checkboxes.

Phase 2 — Plan slices

  • Invoke planner subagent with approved requirements. Optionally use MCP to refine API/library scope. Present slices and the current slice.
  • State — Add one slices/NN-slug.md per slice (copy from slices/01-placeholder.md pattern); update slices/README.md (table + Current slice file); set Current slice file in progress.md; check Phase 2 boxes in progress.md.

Phase 3 — Development cycle per slice (loop)

For each slice until the human says Complete or final approval:

  1. Develop — Implement; use MCP for up-to-date APIs/libs when needed.
  2. Tests — Unit → integration → e2e (or manual checklist) per agile-project-lifecycle/reference.md.
  3. If red — debugger subagent → fix → re-run; treat as a mini /bugfix-and-verify loop.
  4. Verify — verifier subagent (can optionally cross-check with MCP against docs).
  5. Gate 2 — Iteration — Summary + Approve / Next slice / Feedback: … / Complete / Abort. Stop until the human replies.
  6. State — Update only the active slice file for develop/test/verify checkboxes; progress.md snapshot; before Gate 2 stop, Blocked on; after reply, gates.md + progress.md.

Phase 4 — Quality and ship prep

  1. /code-review skill on the changed surface.
  2. If auth/payments/PII touched: security-auditor subagent (readonly).
  3. /deploy-checklist skill before any deploy suggestion.
  4. State — progress.md Phase 4 boxes; prepend sessions.md with outcomes.

Phase 5 — Final release handoff (HITL)

  • Run the same human gate pattern as /release-with-approval: review package, breaking changes, risks; Approved / Request changes / Abort; no tag/deploy unless the user explicitly asks after approval.
  • State — gates.md Gate 3 row; progress.md current phase done; Remaining empty or handoff-only; meta.md timestamp.

Reference

What ships with it: 11 files

8.0 KB alongside SKILL.md

data/

initial-state/

Gives 0 of the 12 instructions most automation workflows skills give in ~1.8k tokens

Counted across 745 of the 1,008 authors here whose files we hold, read 2026-08-07

  • Write conventional commit messagesin 36 of 745, across 35 files
  • Delete branches after mergein 30 of 745, across 21 files
  • Make atomic commitsin 25 of 745, across 15 files
  • Write minimal code to pass testsin 22 of 745, across 10 files
  • Re-snapshot after navigation or DOM changesin 21 of 745, across 13 files
  • Use try-catch for error handlingin 20 of 745, across 8 files
  • Run tests before committingin 20 of 745, across 12 files
  • Write tests before implementationin 20 of 745, across 8 files
  • Configure branch protection rulesin 19 of 745, across 5 files
  • Explain the why in commit messagesin 19 of 745, across 9 files
  • Refactor code while tests remain greenin 19 of 745, across 6 files
  • Interact with elements using refsin 19 of 745, across 11 files

Said here and by no other author read

  • create on-disk state directory from template first
  • read state files in the specified order
  • resume work using progress and gate files
  • update state files after every meaningful step
  • set blocked status in progress file before stopping
  • summarize completed in progress and next work when resuming

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 326,629. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.