🩻 SKELETEN

The AI-First Technical Documentation & Intent Drift Sentry Engine.

SKELETEN is an autonomous documentation framework designed to be deployed directly inside your source repository. It parses abstract syntax architectures using tree-sitter, cross-references the intelligence with vector similarity via sentence-transformers, and actively monitors your intent-to-code gap to identify Documentation Drift before bad code ships.

Crucially, SKELETEN was built for AI Agents. It automatically injects custom skills and slash commands (/skeleten-run) directly into your localized .agents configuration, ensuring that LLMs working in your repository exactly know how to parse context and repair drift safely.

⚡ Quick Start

SKELETEN is distributed globally via NPM and installs flawlessly using a single drop-in command:

npx skeleten init

Running this inside the root of your project immediately translates the python core engine into your repository and configures your virtual environment.

🏗️ The 4 Phases of SKELETEN

1. PHASE_01_CORE: Deep AST extraction using recursive Tree-Sitter parsers. Extracts structural elements (Functions, Classes, Context). Assembles initial JSON matrices.
2. PHASE_02_VECTOR: Sentence Transformer vectorization calculates Cosine Similarities between the raw function names and document tokens. Highlights mismatches.
3. PHASE_03_GEN: Synthesizes formal SKELETEN_API_REFERENCE.md formatted strictly for native Agent consumption with Fenced blocks.
4. PHASE_04_DRIFT: Generates pre-commit hooks that strictly analyze the repository. If Code-to-Intent disparity >15%, commits fail automatically safely guarding production.

🧠 AI Agent Capabilities

Immediately upon running initialization, SKELETEN provisions your .agents configuration path to expose custom workflow hooks. Any compliant contextual AI reading your root can trigger:

• /skeleten-run — Runs the extraction architecture across all 4 phases autonomously.
• /skeleten-health — Tests SKELETEN schema integrity metrics to guarantee zero false positives.

You don't just use SKELETEN. Your AI uses SKELETEN.

🔒 Isolated Execution (venv)

We recognize that Python dependency scopes (requirements.txt) can shatter fragile external mono-repos. We built SKELETEN's extraction engine to autonomously invoke a node child process during execution that guarantees full virtual environment (venv) isolation. run.py dynamically sniffs exec_path strings to force reroutes into .skeleten/venv—acting independently of your root interpreter context.

📜 License

Available underneath standard MIT provisions. Developed cleanly by SSTECH.