agentleFS
mayankjain1815-commitsmayankjain1815-commits/CITADEL/skills/create-app/SKILL.md

create-app

Skill4 starsChanged 4 months agoMITView on GitHub

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.

6 other projects use a version of this, 1 of them edited.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

11 rules. 0 are common, 9 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 5 edited versions we read keep these.

  • 9

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use when the user wants to create an app, add a feature, or scaffold a project.

    line 32
  2. Distinctive:

    If no template matches, fall through to Tier 2.

    line 47
  3. Distinctive:

    Every phase end condition includes "no new typecheck errors" + "existing tests pass"

    line 64
  4. Distinctive:

    Read the user's input. Determine the tier.

    line 84
  5. Very specific:

    Tier 3: Check .planning/_templates/app-types/ for template → present PRD, ask for changes → /architect with template defaults → Archon …

    line 92
  6. Distinctive:

    Run full verification of all PRD end conditions, present results

    line 102
  7. Very specific:

    Read existing codebase — file tree, package.json, key entry points, existing patterns

    line 105
  8. Distinctive:

    Check each PRD end condition (run commands, check files, invoke /live-preview for visual checks).

    line 114
  9. Distinctive:

    Every campaign phase has machine-verifiable end conditions

    line 124
  10. Distinctive:

    Vague requirements: Default to Tier 2 and ask clarifying questions before producing the PRD.

    line 130
  11. Distinctive:

    After the campaign completes and verification runs, output:

    line 140

“Common” is measured: kept by at least 6 in 10 of the 5 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/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 PatternTier
"blank project", "scaffold", "new empty"1
"help me build", "I want to create", "guide me"2
"todo app", "blog", "dashboard", well-known app type3
"build me [detailed]", "create [app]", confident description4
"add [feature]", "implement [feature]", existing project + feature description5
AmbiguousDefault 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:

---HANDOFF---
- App: {name}
- Built: {feature ledger summary}
- Verified: {N}/{total} end conditions passed
- Status: {complete | partial | failed}
- To run: {start command}
- Next: {suggested next step, e.g., /postmortem or deploy command}
- Reversibility: amber -- multi-tier creation, revert the creation commits
---

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.