Skip to content
Back to skills

Side Project Launcher

ASecurity

companyproject basis A pipeline where an agent team collaborates for comprehensive planning. 'companyproject basis', 'company project idea', ' project', 'MVP ', ' reporting ', 'service launch', '1person development', 'indie hacking', 'week project', 'tech stack ', 'company ' etc. companyproject basis before skill usage. existing idea case also tech stack MVP design degreeKRW. However, actual code writing, specialistperson whenplan work, person/from versus is outside this skill's scope.

  • 3 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added May 27, 2026
businessreact

Security analysis

A100/100

Scanned May 27, 2026

npx -y skills add chodolmu/taskforge-pro --skill side-project-launcher --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Side Project Launcher?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Side Project Launcher
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/chodolmu-side-project-launcher/badge)](https://www.skillsdirectory.com/skills/chodolmu-side-project-launcher)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: side-project-launcher
description: "companyproject basis A pipeline where an agent team collaborates for comprehensive planning. 'companyproject basis', 'company project idea', ' project', 'MVP ', ' reporting ', 'service launch', '1person development', 'indie hacking', 'week project', 'tech stack ', 'company ' etc. companyproject basis before skill usage. existing idea case also tech stack MVP design degreeKRW. However, actual code writing, specialistperson whenplan work, person/from versus is outside this skill's scope."
---

# Side Project Launcher — companyproject basis pipeline

ideaverify→tech stack→MVP→developmentroadmap→launchchecklist An agent team collaborates to generate all deliverables at once.

## execution mode

**agent team** — 5 agents communicate directly via SendMessage and cross-verify each other's work.

## agent composition

| agent | file | role | type |
|---------|------|------|------|
| idea-validator | `.claude/agents/idea-validator.md` | market analysis, competition research, differentiation verify | general-purpose |
| techstack-analyst | `.claude/agents/techstack-analyst.md` | tech stack comparison, recommendation, person design | general-purpose |
| mvp-designer | `.claude/agents/mvp-designer.md` | core feature definition, , | general-purpose |
| roadmap-builder | `.claude/agents/roadmap-builder.md` | development schedule, launch strategy, checklist | general-purpose |
| launch-reviewer | `.claude/agents/launch-reviewer.md` | cross-verification, execution possiblenature confirm | general-purpose |

## workflow

### Phase 1: preparation (Orchestrator directly perform)

1. Extract from user input:
 - **idea**: , problem
 - **development competency** (optional): usage possibleKorean technical, development level
 - ** time** (optional): week deploy possible time
 - **budget** (optional): month operations cost Korean
 - **existing file** (optional): basisfrom, etc.
2. `_workspace/` Create the directory at the project root
3. Organize input and save to `_workspace/00_input.md`
4. If existing files are provided, copy them to `_workspace/`and skip the corresponding Phase
5. Determine the **execution mode** based on the scope of the request

### Phase 2: team composition and execution

| order | task | responsible | dependency | deliverable |
|------|------|------|------|--------|
| 1 | idea verify | validator | None | `_workspace/01_idea_validation.md` |
| 2 | tech stack | techstack | task 1 | `_workspace/02_techstack_recommendation.md` |
| 3 | MVP design | designer | task 1, 2 | `_workspace/03_mvp_spec.md` |
| 4 | roadmap·launch | roadmap | task 2, 3 | `_workspace/04_roadmap_launch.md` |
| 5 | launch review | reviewer | task 1~4 | `_workspace/05_review_report.md` |

**teamKRW between flow:**
- validator complete → techstackto project type·difficulty, designerto UVP·differentiation point, roadmapto market deliver
- techstack complete → designerto technical constraint·implementation difficulty, roadmapto learning time·technical risk deliver
- designer complete → roadmapto featureby time·priority deliver
- reviewer all deliverable cross-verification. 🔴 required revision findings when Request revision from the relevant agent -> rework -> re-verify (up to 2 rounds)

### Phase 3: integration and final deliverable

1. `_workspace/` Verify all files in the directory
2. review reportConfirm that all critical revisions from the review report have been addressed
3. Report the final summary to the user

## task scaleby mode

| user request pattern | execution mode | deploy agent |
|----------------|----------|-------------|
| "companyproject overall basis" | **Full pipeline** | 5people beforeKRW |
| " idea verifyplease do" | **verify mode** | validator + reviewer |
| "tech stack recommendationplease do" (idea ) | **technical mode** | validator + techstack + reviewer |
| "MVP " (technical decision) | **MVP mode** | designer + roadmap + reviewer |
| "launch checklist create it" | **launch mode** | roadmap + reviewer |

**existing file utilization**: user basisfrom, tech stack decision etc. provide applicable stage case.

## data deliver protocol

| strategy | method | also |
|------|------|------|
| File-based | `_workspace/` | Store and share major deliverables |
| Message-based | SendMessage | Real-time key information transfer, revision requests |
| Task-based | TaskCreate/TaskUpdate | Progress tracking, dependency management |

File naming convention: `{}_{agent}_{deliverable}.{extensionspecialist}`

## error handling

| error type | strategy |
|----------|------|
| idea un-present | trend based idea 3items proposal after optional also |
| development None | code/code recommendation, learning roadmap included |
| web search failure | general market degreeas analysis, "data limitation" specify |
| agent failure | Retry once -> proceed without that deliverable, note the gap in the review report |
| reviewfrom 🔴 findings | Request revision from the relevant agent -> rework -> re-verify (up to 2 rounds) |

## test scenario

### flow
**Prompt**: "developmentspecialist interview preparation week web companyproject . React week developmentto do ."
**expected result**:
- verify: interview preparation market analysis, existing service(LeetCode, ) versus differentiation
- technical: React + Supabase + Vercel recommendation, free utilization
- MVP: core feature 3~5items, , data model
- roadmap: 4week sprint, Product Hunt launch strategy
- review: overall consistency verify, risk matrix

### existing file utilization flow
**Prompt**: "un- idea verify tech stack also . MVP only " + basisfrom department
**expected result**:
- MVP mode: designer + roadmap + reviewer deploy
- validator, techstack case

### error flow
**Prompt**: " reporting idea "
**expected result**:
- validator user company question or trend based idea 3items proposal
- optional after Full pipeline progress
- review report "idea stage" specify

## agentby extension skill

| agent | extension skill | also |
|---------|----------|------|
| idea-validator | `market-sizing-calculator` | TAM/SAM/SOM calculation, revenue model verify |
| techstack-analyst | `techstack-decision-matrix` | tech stack comparison matrix, person cost total |

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.