magician

       *       
      /|\      
     / | \     
    /  *  \    
   /_______\   
  /^  ^.^  ^\  
  \  ~(u)~  /  
   \_______/~  
     |   |  /  

What it does

Most AI coding tools require you to describe your stack, pick templates, and manage context manually. Magician inspects your project on every session start, assembles targeted knowledge for every technology it finds, and gets smarter with every session.

One command to go from idea to PR:

/manifest

How it works

The manifest flow — full autonomous SDLC

flowchart TD
    A["/manifest"] --> B{"scope OK?"}
    B -- too large --> C["decompose\ninto sub-projects"]
    B -- ok --> D["/conjure\ndesign dialogue"]
    D --> E["approved spec"]
    E --> F["/blueprint\nimpl plan + parallelism map"]
    F --> G["/portal\ngit worktree isolation"]
    G --> H["/orchestrate\nparallel agents"]
    H --> I["/ward\nTDD throughout"]
    I --> J["/certify\nverify: tests + browser"]
    J --> K{all green?}
    K -- no --> H
    K -- yes --> L["/scrutinize\nreview + remediate"]
    L --> N["/seal\nPR + loop until merged"]

    style A fill:#6c63ff,color:#fff
    style D fill:#6c63ff,color:#fff
    style F fill:#6c63ff,color:#fff
    style H fill:#43e97b,color:#000
    style I fill:#43e97b,color:#000
    style J fill:#43e97b,color:#000
    style L fill:#43e97b,color:#000
    style N fill:#4facfe,color:#000

Human gates (4 only): scope confirm → spec approval → plan approval → PR title. Everything else: autonomous.

Dynamic project inspector — no manual stack selection

flowchart LR
    A["session start"] --> B["scan project files"]
    B --> C{"detect markers"}

    C --> D["package.json\ntsconfig.json\nnext.config.*"]
    C --> E["pom.xml\nbuild.gradle\n*.xcodeproj"]
    C --> F["go.mod\nCargo.toml\npyproject.toml"]
    C --> G["pubspec.yaml\ncapacitor.config.*\nproject.godot"]

    D --> H["assemble lore\nweb/ fragments"]
    E --> I["assemble lore\nbackend/ fragments"]
    F --> J["assemble lore\nbackend/ fragments"]
    G --> K["assemble lore\nmobile/ craft/ fragments"]

    H --> L["assign archetype\n+ inject context"]
    I --> L
    J --> L
    K --> L

    L --> M["session ready\nin < 2 seconds"]

    style A fill:#0d1117,color:#ccc,stroke:#555
    style M fill:#43e97b,color:#000

Polyglot stacks (Next.js + FastAPI + Go) get full coverage automatically. No pack selection needed.

Self-learning — intelligence grows each session

flowchart TD
    A["session ends"] --> B["chronicle-stop.sh\nStop hook"]
    B --> C["git log + diff\nobservable data only"]
    C --> D["write chronicle entry\nCLAUDE_PLUGIN_DATA/chronicle/"]
    D --> E["update patterns.json"]

    E --> F{"pattern seen 3x?"}
    F -- yes --> G["offer: create skill\nvia /inscribe"]
    F -- no --> H[" "]

    G --> I["next session start"]
    H --> I

    I --> J["load last 3 entries\nas additionalContext"]
    J --> K["cumulative intelligence\ngrows without replay"]

    style B fill:#f7971e,color:#000
    style D fill:#f7971e,color:#000
    style G fill:#43e97b,color:#000
    style K fill:#6c63ff,color:#fff

Skills

18 skills. Each carries modern frontmatter (allowed-tools to cut permission prompts in auto mode, disable-model-invocation on side-effecting standalone skills, argument-hint, context: fork for heavy read-only work) and scales reasoning effort to the task.