Your prompt
    │
    ▼
┌──────────────────────────────────────────────┐
│  Security scan — block injections first      │
└──────────────────────────────────────────────┘
    │
    ▼
┌──────────────────────────────────────────────┐
│  Pillar 1 — Skills                           │
│  Score 1,944 skills against your task        │
│  Inject top 3 + all 32 instincts             │
│  (instincts fire every single prompt)        │
└──────────────────────────────────────────────┘
    │
    ▼
┌──────────────────────────────────────────────┐
│  Pillar 2 — Memory                           │
│  Load decisions, patterns, errors            │
│  for this specific project                   │
│  Record accept/rework signals after          │
└──────────────────────────────────────────────┘
    │
    ▼
┌──────────────────────────────────────────────┐
│  Pillar 3 — Agents (complex tasks only)      │
│  Decompose → spawn specialists               │
│  Frontend agent gets UI skills               │
│  Backend agent gets API skills               │
│  Nobody else does this                       │
└──────────────────────────────────────────────┘
    │
    ▼
Claude (or any AI tool)

# macOS (Apple Silicon)
curl -L https://skillgod.dev/dist/sg-mac-arm64 -o /usr/local/bin/sg && chmod +x /usr/local/bin/sg

# macOS (Intel)
curl -L https://skillgod.dev/dist/sg-mac-intel -o /usr/local/bin/sg && chmod +x /usr/local/bin/sg

# Linux
curl -L https://skillgod.dev/dist/sg-linux -o /usr/local/bin/sg && chmod +x /usr/local/bin/sg

# Windows (PowerShell)
irm https://skillgod.dev/install.ps1 | iex

# From source
git clone https://github.com/amancodingrepo/skillgod
cd skillgod
pip install -r requirements.txt
python engine/ingest.py
sg init

sg init                          # detects IDE, writes .mcp.json, rebuilds index
sg find "debug python traceback" # test skill discovery
sg stats                         # vault counts and health
sg scan "your prompt here"       # test security scanner

sg init              Set up SkillGod — detects IDE, writes .mcp.json
sg find <task>       Score vault against a task, show top matches
sg stats             Vault health, category counts, index status
sg build             Interactive skill builder — guided prompts
sg learn             Save any output as a new skill (goes to meta/)
sg sync              Rebuild local index (free) or decrypt full vault (pro: sg sync --key YOUR-KEY)
sg scan <text>       Scan for prompt injection threats
sg signals           Accept/rework rates per skill — see what's working
sg promote           Review auto-learned skills, approve for vault promotion
sg run <tool>        Run any AI tool with SKILLGOD_ROOT injected

find_skills(task)         → top scored skills for this task
inject_context(task)      → full augmented prompt string
save_memory(summary)      → persist a decision to project memory
get_memory(project)       → retrieve recent project context
learn_skill(task, output) → auto-generate skill from session output
stocktake()               → vault health report
spawn_agents(task)        → decompose and route to specialist agents
security_scan(text)       → threat detection
vault_stats()             → live vault counts

skillgod/
├── engine/
│   ├── ingest.py        ← sources → vault normaliser
│   ├── memory.py        ← SQLite memory layer
│   ├── skills.py        ← scoring + injection engine
│   ├── agents.py        ← spawning + per-agent skill routing
│   ├── runtime.py       ← combines all three pillars
│   ├── mcp_server.py    ← FastMCP server on :3333
│   ├── security.py      ← injection detection
│   ├── encryption.py    ← AES-256-GCM vault encryption
│   ├── signals.py       ← accept/rework tracking
│   └── variants.py      ← promotion queue
├── hooks/
│   ├── session_start.py ← loads instincts + memory
│   ├── pre_tool.py      ← security scan + skill injection
│   └── post_tool.py     ← captures memory + learns skills
├── vault_free/          ← 30 free skills (full vault unlocked with Pro)
└── cli/                 ← Go binary: sg