Install
$ agentstack add skill-vladislavsournine-vladyslav-skills-init-project ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ● Environment & secrets Used
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
Init Project
Type: Engineer (light)
Overview
Bootstrap a new project with the minimal Claude Code shell (CLAUDE.md, .gitignore, .claude/settings.json). In interactive mode an opt-in menu adds documentation, backend infrastructure, and agent stubs — only the modules you actually tick. Everything runs via focused bash modules under scripts/modules/; no LLM cost for the mechanical scaffold work.
Process
Step 0: Pre-flight (Opus main)
All Q&A happens here before any bash is executed.
- Verify working directory. Run
ls -Aon the target directory (default: cwd). If it is non-empty — anything beyond.git/alone — ask via AskUserQuestion:
> "The current directory is not empty. Contents: ``. Continue and add project structure here, or abort?"
- Abort → exit cleanly. Do not proceed.
- Continue → proceed. Modules skip files that already exist.
- AskUserQuestion — project name (free text):
> "What is the project called?"
- AskUserQuestion — mode (single-select):
> "Scaffold mode?"
- Options:
minimal|interactive minimal— only core module runs (bare AI shell).interactive— core runs first, then the module menu (Step 2) is presented.
- Resolve plugin root.
- Glob
~/.claude/plugins/cache/vladyslav-marketplace/vladyslav/*/scripts/modules/core.sh. - Take the matching path and go three levels up (strip
scripts/modules/core.sh) to get ``. - Fall back to
/Volumes/DevSSD/Development/vladyslav-skillsif Glob returns nothing (development clone). - Verify
/skills/init-project/assets/exists. If missing, warn the user and continue — assets are optional for core operation.
Capture: `, , , `.
Step 1: Core (always)
Execute via Bash tool:
/scripts/modules/core.sh \
--pwd \
--plugin-root \
--name
The script emits a single line of JSON on stdout:
{
"status": "success" | "partial" | "error",
"files_written": ["", ...],
"files_skipped": ["", ...],
"warnings": ["", ...],
"error": ""
}
Capture this JSON. If status is error, surface the error and stop — do not proceed to Step 2.
Step 2: Module menu (interactive mode only)
Skip entirely if ` is minimal`.
Present three grouped AskUserQuestion multi-selects, each with all options unticked by default. Run only the modules the user ticks. Unticked items are silently skipped — "don't know → don't create".
Group A — Docs
> "Which documentation scaffolds do you want?" > (all unticked by default)
| Option | Module call | |---|---| | prd + planning | docs.sh --pwd --plugin-root | | design system | design-system.sh --pwd --plugin-root --name | | architecture | architecture.sh --pwd --plugin-root |
Group B — Backend infra
> "Which backend infrastructure do you want?" > (all unticked by default)
| Option | Module call | |---|---| | docker | docker.sh --pwd --plugin-root | | postgres | postgres.sh --pwd --plugin-root | | redis | redis.sh --pwd --plugin-root | | alembic | alembic.sh --pwd --plugin-root | | backend skeleton | backend-skeleton.sh --pwd --plugin-root |
Group C — Agents
> "Which Claude Code agents do you want installed?" > (all unticked by default) > > Options: architect-reviewer, backend-engineer, qa-reviewer, release-manager
Collect the ticked names into a comma-separated string, then run:
/scripts/modules/agents.sh \
--pwd \
--plugin-root \
--agents ""
If no agents are ticked, skip the agents.sh call entirely.
Each module call emits the same JSON shape as core. Capture all outputs.
Step 3: Roadmap gate
After modules complete, ask via AskUserQuestion:
> "Які ключові фічі плануєш в цьому проекті? Розіб'ємо на MVP-фази в ROADMAP.md"
- If the user provides features → generate
ROADMAP.mdat/ROADMAP.mdusing the phase structure below, then commit the file to git. - If the user skips ("потім", "не знаю", empty answer, or declines) → do not create the file. Continue normally.
This step is non-blocking — init-project completes regardless of the answer.
ROADMAP.md phase structure:
# Roadmap:
> Created: YYYY-MM-DD
## Phase 1:
**Done when:**
- [ ] Task 1
- [ ] Task 2
- [ ] Task 3
## Phase 2:
**Done when:**
- [ ] Task 1
- [ ] Task 2
Phases represent MVP milestones — what ships in the first release vs. what comes after.
Step 4: Summary + smart Next
Merge all module JSON outputs. Render to the user:
✓ init-project complete
Project: in
Mode:
Files written:
Skipped (already existed):
Warnings:
ROADMAP.md:
Then append a context-aware Next: line — pick the first that applies:
| Condition | Next line | |---|---| | No docs module was ticked | Next: docs народяться за потреби — discover / write-user-stories / write-test-docs. | | Any backend-infra module was ticked | Next: /vladyslav:add-feature щоб почати фічу. | | minimal mode | Next: повернись у interactive за потреби, або одразу /vladyslav:add-feature. |
The skill never auto-runs the next skill. The Next line is informational only.
status: error from any module: surface the error and which module failed. List files written before the failure. Do not retry automatically.
Why this is a Light Engineer skill
- No subagent dispatch. Each module is a deterministic bash script. No LLM reasoning is needed for file creation.
- Modular and opt-in. Only requested modules run, so a minimal project stays minimal.
- Idempotent. Modules skip files that already exist and report them in
files_skipped. - ~1 second, 0 LLM tokens per module. Q&A in Step 0 is the only model-driven part.
Output allowlist
Paths the modules may write (relative to `). The modules never modify pre-existing files except where noted (append operations on docker-compose.yml and .env`).
core.sh (always):
.gitignore
.claude/settings.json
CLAUDE.md
.remember/now.md
docs.sh:
docs/product/prd.md
docs/plans/tasks.md
docs/plans/backlog-next.md
design-system.sh:
docs/design/system.md
architecture.sh:
docs/architecture/system.md
docker.sh:
Dockerfile
docker-compose.yml
docs/operations/docker.md
postgres.sh:
docker-compose.yml (appends postgres service block)
.env (appends DATABASE_URL)
redis.sh:
docker-compose.yml (appends redis service block)
.env (appends REDIS_URL)
alembic.sh:
alembic.ini
migrations/env.py
migrations/versions/.gitkeep
migrations/README
backend-skeleton.sh:
requirements.txt
src/__init__.py
src/main.py
agents.sh:
.claude/agents/.md (one per chosen agent)
Roadmap gate (Step 3):
ROADMAP.md
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: VladislavSournine
- Source: VladislavSournine/vladyslav-skills
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.