create-app
End-to-end app creation from a single description.
Skill by mayankjain1815-commits
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/mayankjain1815-commits-create-app
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: marketing
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/create-app
What's inside
- Skill name
create-app- Description
- End-to-end app creation from a single description. Five tiers: blank project, guided, templated, fully generated, or feature addition to existing codebase. Routes through PRD, architecture, and Archon campaign with verification at every step.
- License
- MIT
- Instructions
- 132 lines in SKILL.md · view on GitHub
Files (3)
- SKILL.md
- existing-codebase.md · __benchmarks__/
- greenfield-vague.md · __benchmarks__/
SKILL.md (first lines)
# /create-app — From Description to Verified Application
## Orientation
**Use when:** building a new app from scratch -- full scaffold, design system, and feature set from a single description.
**Don't use when:** adding a feature to an existing app (use /marshal or /archon); generating a single component (use /scaffold).
## Tier Detection
Use when the user wants to create an app, add a feature, or scaffold a project. Classify the input into one of five tiers:
### Tier 1: Blank Project
- Trigger: "create a blank project", "new project", "scaffold"
- Action: Run /scaffold with stack detection. No PRD, no architecture.
### Tier 2: Guided
- Trigger: "I want to build...", "help me create...", description with questions
- Action: /prd → user approves → /architect → user approves → /archon
- Human checkpoints: after PRD, after architecture, before each major phase.
### Tier 3: Templated
- Trigger: Describes a well-known app type ("a todo app", "a blog", "a dashboard with auth")
- Action: Load template PRD if available → /architect with template defaults → /archon
- Template detection: check `.planning/_templates/app-types/` for matching templates.
If no template matches, fall through to Tier 2.
### Tier 4: Generated (Full Autonomy)
- Trigger: "build me [detailed description]", "create [app] and deploy it"
- Action: /prd (minimal questions) → /architect (auto-approve if confidence high) → /archon with self-correction loop
- Human checkpoints: after PRD only. Architecture and execution are autonomous.
- Safety: all Archon self-correction mechanisms active. Direction alignment every
2 phases. Quality spot-checks every phase. Circuit breakers armed.
### Tier 5: Feature Addition (Existing Codebase)
- Trigger: existing project + feature description ("add auth", "add a dashboard", "add dark mode")
- Detection: project has source files (src/, app/, lib/, package.json with deps) AND description is a feature, not a standalone app
- Action: /prd in feature mode → /architect in existing codebase mode → /archon
- Key differences from greenfield tiers:
- PRD reads existing codebase before asking questions
- Architecture describes changes to existing files, not a standalone system
- Phase 0 is always "Baseline" — record current typecheck/test state
- Every phase end condition includes "no new typecheck errors" + "existing tests pass"
- Risk register always includes "regression in existing functionality"
- Human checkpoints: after feature spec (PRD). Architecture can auto-approve if
the feature is well-scoped and all conditions are machine-verifiable.
### Tier Classification
| Input Pattern | Tier |
|---|---|
| "blank project", "scaffold", "new empty" | 1 |
| "help me build", "I want to create", "guide me" | 2 |
| "todo app", "blog", "dashboard", well-known app type | 3 |
| "build me [detailed]", "create [app]", confident description | 4 |
| "add [feature]", "implement [feature]", existing project + feature description | 5 |
| Ambiguous | Default to Tier 2 (safest) |
## Protocol
### Step 1: CLASSIFY
Read the user's input. Determine the tier. Announce what you'll do in plain language — not tier numbers. If the classification is wrong, accept user overrides ("just scaffold it" → Tier 1, "just build it" → Tier 4).
### Step 2: EXECUTE TIER
**Tier 1:** Invoke /scaffold. Done.
**Tier 2:** /prd → user approves → /architect → user approves → Archon campaign → brief user after each major phase.
**Tier 3:** Check `.planning/_templates/app-types/` for template → present PRD, ask for changes → /architect with template defaults → Archon campaign.
**Tier 4:**
1. /prd express mode (0-1 questions)
2. User approves PRD (only mandatory checkpoint)
3. /architect (auto-approve if all end conditions are machine-verifiable)
4. Archon campaign with ALL safety systems active:
- Direction alignment every 2 phases, quality spot-check every phase
- Circuit breakers: 3 failures = new approach, 5+ type errors = park
5. Execute autonomously until complete or parked
6. Run full verification of all PRD end conditions, present results
**Tier 5 (Feature Addition):**
1. Read existing codebase — file tree, package.json, key entry points, existing patterns
2. /prd in feature mode (max 2 questions)
3. User approves feature spec (one mandatory checkpoint)
4. /architect in existing codebase mode — Phase 0 always "Baseline" (run typecheck + tests, record counts). Auto-approve if machine-verifiable.
5. Archon campaign — every phase end condition includes "no new typecheck errors vs baseline" and "existing tests pass"
6. On completion: verify all feature end conditions + baseline regression check, present results
### Step 3: VERIFY (All Tiers except 1)
Check each PRD end condition (run commands, check files, invoke /live-preview for visual checks). Report PASS / PARTIAL / FAIL with specifics.
### Step 4: DELIVER
Present: what was built, what was verified, what needs attention, how to run it, and suggested next step (e.g., /postmortem or deploy command).
## Quality Gates
- PRD exists and is approved before any code is written
- Architecture exists before Archon starts
- Every campaign phase has machine-verifiable end conditions
- Final verification checks all PRD end conditions
- User receives a clear report of what was built and what needs attention
## Fringe Cases
**Vague requirements**: Default to Tier 2 and ask clarifying questions before producing the PRD.
**Project already initialized**: Existing source files (src/, app/, package.json with deps) → automatically classify as Tier 5. Do not scaffold over an existing project.
**If .planning/ does not exist**: /prd and /architect will create it. If not possible, present inline and ask the user to run `/do setup` first.
**Tier misclassification**: Switch immediately on user correction without re-reading the input.
## Exit Protocol
After the campaign completes and verification runs, output:From mayankjain1815-commits/CITADEL, MIT.
More from mayankjain1815-commits/CITADEL
- 0.519score
daemon
mayankjain1815-commits
Continuous autonomous operation mode.
SkillCFreshness: C★ 4Claude Code +9 - 0.519score
grill
mayankjain1815-commits
Relentless one-question-at-a-time interview that sharpens a vague plan or design into shared understanding before any build.
SkillCFreshness: C★ 4Claude Code +9 - 0.519score
decision-map
mayankjain1815-commits
Turn a loose idea into a git-tracked, session-resumable map of typed investigation tickets, then drive them to resolution one at a time.
SkillCFreshness: C★ 4Claude Code +9 - 0.519score
evolve
mayankjain1815-commits
Research-driven multi-cycle improvement director.
SkillCFreshness: C★ 4Claude Code +9 - 0.519score
create-skill
mayankjain1815-commits
Creates new skills from the user's repeating patterns.
SkillCFreshness: C★ 4Claude Code +9 - 0.519score
fleet
mayankjain1815-commits
Parallel campaign orchestrator.
SkillCFreshness: C★ 4Claude Code +9 - 0.516score
design
mayankjain1815-commits
Generates and maintains a design manifest for visual consistency.
SkillCFreshness: C★ 4Claude Code +9 - 0.516score
do
mayankjain1815-commits
Unified router that auto-routes user intent to the right orchestrator or skill.
SkillCFreshness: C★ 4Claude Code +9 - 0.516score
doc-gen
mayankjain1815-commits
Documentation generator with three modes: function-level (JSDoc/docstrings), module-level (directory READMEs), and API reference (endpoints/exports).
SkillCFreshness: C★ 4Claude Code +9 - 0.516score
experiment
mayankjain1815-commits
Automated optimization loop with scalar fitness function.
SkillCFreshness: C★ 4Claude Code +9 - 0.516score
architect
mayankjain1815-commits
Given a PRD, produces an implementation architecture: file tree, component breakdown, data model, and a phased build plan with end conditions that Archon can…
SkillCFreshness: C★ 4Claude Code +9 - 0.516score
archon
mayankjain1815-commits
Autonomous multi-session campaign agent.
SkillCFreshness: C★ 4Claude Code +9
Signals
- Stars
- 4
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Jun 11, 2026 (4 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
Same skill elsewhere
Copies of this SKILL.md published by others; search shows only one of them.
More from mayankjain1815-commits
- 0.519score
evolve
mayankjain1815-commits
Research-driven multi-cycle improvement director.
SkillCFreshness: C★ 4Claude Code +9 - 0.519score
fleet
mayankjain1815-commits
Parallel campaign orchestrator.
SkillCFreshness: C★ 4Claude Code +9 - 0.519score
daemon
mayankjain1815-commits
Continuous autonomous operation mode.
SkillCFreshness: C★ 4Claude Code +9 - 0.519score
decision-map
mayankjain1815-commits
Turn a loose idea into a git-tracked, session-resumable map of typed investigation tickets, then drive them to resolution one at a time.
SkillCFreshness: C★ 4Claude Code +9 - 0.519score
create-skill
mayankjain1815-commits
Creates new skills from the user's repeating patterns.
SkillCFreshness: C★ 4Claude Code +9 - 0.519score
grill
mayankjain1815-commits
Relentless one-question-at-a-time interview that sharpens a vague plan or design into shared understanding before any build.
SkillCFreshness: C★ 4Claude Code +9
Cite this page
create-app skill. Agent Mart. https://agentmart.dev/skills/mayankjain1815-commits-create-app (last updated 2026-10-24)