| title | jev-superpowers: Systematic Agentic Skills with TypeSafe Jev & Laya | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| description | Systematic software development framework for AI coding agents upgraded with TypeSafe Jev and open-weight Laya System One typed decisions, zero-hallucination package vetting, and completion gates. | |||||||||
| canonical | https://github.com/AkashPriyadarshii/jev-superpowers | |||||||||
| keywords |
|
Support: fuel the next build —
Systematic agentic software development upgraded with TypeSafe Jev System One typed decisions & open-weight Laya local inference.
Why • Core Superpowers • Dual Backend: Cloud vs Local • The Toolchain Suite • Quickstart • Usage Guide • Architecture • Non-Goals • Ecosystem
OpenJEV support: Jev is built by TypeSafe. This fork keeps TypeSafe as the default and adds optional support for OpenJEV, a free community gateway to the same Jev model — set OPENJEV_API_KEY (or JEV_PROVIDER=openjev) to use it. Original project: https://github.com/AkashPriyadarshii/jev-superpowers by @AkashPriyadarshii.
Autonomous coding agents equipped with standard development methodologies still fail in predictable, expensive ways:
- Hallucinated Packages: Agents invent fake npm packages or dead crates during implementation planning.
- Arbitrary Architecture Guessing: Agents flip coins on stack decisions and database choices during brainstorming without calibrated confidence.
- Unchecked Destructive Execution: Agents run destructive shell commands or push unverified git commits containing credentials or bugs.
- Premature Turn Completion: Agents declare tasks "done" based on optimistic assumptions rather than verified evidence.
jev-superpowers upgrades the battle-tested obra/superpowers methodology with System One non-autoregressive decision models. Every architectural fork, dependency selection, shell command, and completion check is gated by sub-second deterministic mathematical judgments—available via TypeSafe AI Cloud (jev-1.13.0) or 100% offline via Laya (convaiinnovations/laya, Apache-2.0).
Design targets measured on the maintainer setup; reproduce with scripts/test.sh plus your own Jev/Laya usage logs before quoting them:
| Metric | Standard Agent Superpowers | TypeSafe Cloud Jev Upgrade | Local FOSS Laya (421M) Upgrade |
|---|---|---|---|
| Package Hallucination Rate | ~14% unvetted libraries | near 0% (jev-scout) |
near 0% (jev-scout offline) |
| Architectural Gate Latency | 3.5s – 12s (generative LLM) | ~70ms – 120ms | ~33ms – 38ms (GPU/Apple Silicon) |
| Per-Decision Verification Cost | ~$0.015 – $0.030 | ~$0.00001 ($0.042/Mtok) | $0.00000 (100% Free / Self-hosted) |
| Data Privacy | Cloud LLM prompt logging | Ephemeral cloud evaluation | 100% Air-Gapped Local (Zero egress) |
| Pre-Commit Diff Screening | Manual / none | Sub-second automated gate | Sub-second automated gate |
| Local Test Suite Run | N/A | 24/24 passed offline | 24/24 passed offline |
jev-superpowers is completely backend-agnostic. Choose the execution profile that fits your operational requirements:
┌────────────────────────┐
│ jev-superpowers │
│ (git-jev, limpet, axi) │
└───────────┬────────────┘
│
┌───────────────────────┴───────────────────────┐
▼ ▼
[Option A: Cloud Jev] [Option B: Local FOSS Laya]
• TypeSafe API (jev-latest) • Open-weight ModernBERT-large (421M)
• Zero local RAM usage • Apache-2.0 License
• export TYPESAFE_API_KEY="..." • 100% Offline / Zero cloud calls
• export TYPESAFE_BASE_URL="http://127.0.0.1:8000"
For complete local setup instructions, see the 100% Local FOSS Engine Guide (docs/FOSS_LAYA.md).
| Skill | Phase | Standard Superpower |
jev-superpowers Upgrade |
|---|---|---|---|
jev-brainstorming |
Ideation | brainstorming |
jev-axi pick trade-off convergence with calibrated confidence ( |
jev-writing-plans |
Planning | writing-plans |
jev-scout zero-hallucination crate & repository verification |
jev-executing-plans |
Execution | executing-plans |
jev-guard command safety + git-jev pre-commit reflex gate |
jev-systematic-debugging |
Triage | systematic-debugging |
jev-axi triage error analysis + Jev Score hypothesis ranking |
jev-verification |
Completion | verification-before-completion |
limpet turn stop-hook + supercov quality anti-pattern scoring |
jev-superpowers orchestrates dedicated single-purpose tools powered by System One decision engines:
| Tool | Role | Install Command | Repository / Package |
|---|---|---|---|
git-jev |
Sub-second Git pre-commit & pre-push reflex gate | Precompiled binary / cargo install |
AkashPriyadarshii/jev-git |
jev-scout |
Zero-hallucination crate & repo scout | cargo install jev-scout |
AkashPriyadarshii/jev-scout |
jev-axi |
Fast System One diff review, error triage & choice pick | npm install -g jev-axi |
shiftynick/jev-axi |
jev-guard |
Destructive shell command filter (<80ms) | Installed with jev-axi / standalone |
leepokai/jev-guard |
supercov |
Jev code quality & anti-pattern oracle | npm install -g supercov |
supercorp-ai/supercov |
limpet |
Agent stop-hook completion policy gate | git clone https://github.com/noplan-inc/limpet / Claude plugin |
noplan-inc/limpet |
jev-seo |
Zero-cost SEO/GEO audit radar & citation scoring | cargo install jev-seo |
AkashPriyadarshii/jev-seo |
For detailed recipes and terminal execution traces of each gate, see the Full Usage & Recipes Guide (docs/USAGE.md).
git clone https://github.com/AkashPriyadarshii/jev-superpowers.git
cd jev-superpowers# 1. Export your free API key from https://console.typesafe.ai
export TYPESAFE_API_KEY="your_api_key"
# 2. Run cross-platform installer
bash install.sh# 1. Install Laya
pip install laya
# 2. Start the local System 1 bridge
python scripts/serve-laya.py --port 8000 &
# 3. Export local routing (Zero cloud tokens, 100% air-gapped)
export TYPESAFE_BASE_URL="http://127.0.0.1:8000"
export TYPESAFE_API_KEY="local"
# 4. Run installer
bash install.sh# 1. Get your free API key from https://openjev.sh/dashboard
export OPENJEV_API_KEY="your_api_key"
# 2. Run cross-platform installer
bash install.shSet JEV_PROVIDER=openjev to force the OpenJEV gateway even when TYPESAFE_API_KEY is also present.
usage: serve-laya.py [-h] [--port PORT] [--host HOST]
Local FOSS System 1 Server (Laya / ModernBERT)
options:
-h, --help show this help message and exit
--port PORT Port to bind (default 8000)
--host HOST Host address (default 127.0.0.1)
Run the offline test suite (bash scripts/test.sh or pwsh scripts/test.ps1):
🧪 Running jev-superpowers offline verification suite...
✔ jev-using-superpowers: valid frontmatter
✔ jev-brainstorming: valid frontmatter
✔ jev-writing-plans: valid frontmatter
✔ jev-executing-plans: valid frontmatter
✔ jev-systematic-debugging: valid frontmatter
✔ jev-verification: valid frontmatter
✔ Cross-platform installers verified (install.sh, install.ps1)
✔ hooks.json wires Stop gate
✔ hooks.json wires PreToolUse gate
✔ pre-commit blocks on git-jev FAIL
✔ stop gate executable
✔ pre-commit ignores plain ls
✔ session-start injects Jev router
✔ installer fails without key
✔ jev-using-superpowers documents failure modes
✔ jev-brainstorming documents failure modes
✔ jev-writing-plans documents failure modes
✔ jev-executing-plans documents failure modes
✔ jev-systematic-debugging documents failure modes
✔ jev-verification documents failure modes
✔ confidence policy exists
✔ FOSS Laya docs exist
✔ serve-laya syntax valid
✔ README avoids absolute claims
Test results: 24 passed, 0 failed.
jev-superpowers/
├── .github/workflows/test.yml # Automated offline validation
├── docs/PRD.md # Product requirements
├── docs/DESIGN.md # System design & hook contracts
├── docs/ARCHITECTURE.md # Architecture dataflow
├── docs/FOSS_LAYA.md # 100% Local FOSS System 1 guide
├── docs/HANDOFF.md # Maintenance guide
├── docs/USAGE.md # Usage guide & recipes
├── memory/DECISIONS.md # Jev decision log
├── scripts/serve-laya.py # Local FOSS System 1 server bridge
├── scripts/test.ps1 # Windows test harness
├── scripts/test.sh # Unix test harness
├── site/ # Documentation marketing site
├── skills/jev-using-superpowers/ # Root router
├── skills/jev-brainstorming/ # Ideation & Jev convergence
├── skills/jev-writing-plans/ # Planning & crate verification
├── skills/jev-executing-plans/ # Execution & safety shield
├── skills/jev-systematic-debugging/ # Error triage & root cause
├── skills/jev-verification/ # Completion gate
├── install.ps1 # Windows installer
├── install.sh # Linux/macOS installer
├── CLAUDE.md # Repo guidelines
├── AGENTS.md # Agent rules
├── STATE.md # Milestone tracker
├── CHANGELOG.md # Release log
├── LICENSE # MIT License
└── README.md # Documentation
- Not a chat agent: Does not provide a conversational chatbot; strictly provides structured agent skills.
- No forced cloud telemetry: Operates 100% offline via local CLIs and Laya open-weight engine when configured.
- No multi-agent swarm bloat: Zero heavy orchestration frameworks or Docker containers required.
- jev-seo — FOSS zero-cost SEO & GEO search radar with TypeSafe Jev
- jev-curate — Jev-curated content toolkit
- jev-git — Sub-second Git pre-commit & pre-push semantic reflex gate powered by TypeSafe Jev
- tdlib-android — Precompiled TDLib native binaries for all 4 Android ABIs
- kharcha — India-first offline-first UPI expense tracker for Android
Akash Priyadarshi
Patna, Bihar, India
- GitHub: @AkashPriyadarshii
- Portfolio: akashpriyadarshi.vercel.app
- LinkedIn: Akash Priyadarshi
- Resume: akashpriyadarshii.github.io/Resume
Social: X / Twitter • Threads • Instagram • Reddit
Built with TypeSafe AI System One & Open-Weight Laya for zero-hallucination autonomous engineering.