Open Source Toolkit
Skillful-Alhazen
The name
Ibn al-Haytham (c. 965–1040), known in the West as Alhazen, was a 10th century Arabic genius who pioneered the scientific method and discovered the optics of the eye. Working in Cairo, he insisted that claims about the natural world must be tested through systematic observation and experiment — not accepted on authority alone.
Skillful-Alhazen, the toolkit, carries the same spirit: don’t just retrieve answers — systematically investigate and structure the evidence.
Architecture
Skillful-Alhazen is built around three layers:
Agent layer
Claude Code orchestrates the curation pipeline — reading documents, calling skills, managing workflow, and reasoning about results.
Knowledge layer
TypeDB provides a typed, schema-enforced knowledge graph. Entities, relationships, and rules — not just vectors and embeddings.
Skills layer
Modular skill modules handle specific tasks — search APIs, PDF parsing, entity extraction, graph queries, report generation.
Skills
Skills are modular capabilities that agents invoke during curation. Each skill wraps a specific function — an API call, a parsing routine, a graph operation.
| Skill | Category | Description |
|---|---|---|
| scientific-literature | Literature | Multi-source scientific literature search and ingestion — Europe PMC, PubMed, OpenAlex, bioRxiv/medRxiv — with semantic search and thematic clustering via Voyage AI + Qdrant. |
| literature-trends | Literature | Abductive argumentation-based trend analysis that traces how explanatory hypotheses evolve over time within a tagged literature thread. |
| jobhunt | Career Intelligence | Track job applications, analyze positions, identify skill gaps, and plan a job search strategy using a TypeDB knowledge graph. |
| techrecon | Technology | Systematically investigate external software systems, libraries, and frameworks — ingesting repos, extracting architecture, and building an understanding graph. |
| alg-precision-therapeutics | Biomedical | Investigate rare disease mechanisms of harm and therapeutic strategies from a known MONDO diagnosis, linking genes, pathways, phenotypes, and drugs. |
| typedb-notebook | Core | Remember, recall, organize, and query knowledge in the Alhazen TypeDB knowledge graph — the persistent memory layer shared by all skills. |
| web-search | Core | Web search via a self-hosted SearXNG metasearch engine aggregating Google, DuckDuckGo, Bing, and other engines — no API key required. |
| domain-modeling | Core | Design and implement new domain-specific knowledge skills using the 6-phase curation pattern (Foraging → Ingestion → Sensemaking → Analysis → Reporting). |
Autonomous deployment
For production use, Skillful-Alhazen can be deployed as a fully autonomous agent via OpenClaw — a hardened deployment of Claude Code with a Squid proxy, MCP server, and Telegram interface running on a Mac Mini or VPS.
Local development
Claude Code running directly on your machine — full toolkit access, interactive workflow, fast iteration.
OpenClaw production
Autonomous deployment on a Mac Mini or Linux VPS — containerized, proxied, and reachable via Telegram for always-on agentic curation.
OpenClaw enables long-running curation workflows that operate without a human in the loop — ingesting new papers overnight, monitoring job postings, or scanning technical landscapes as they evolve.
Getting started
Skillful-Alhazen is open-source and under active development. The project is split into two repositories:
- skillful-alhazen — the core library
- alhazen-skill-examples — example skills that use the system
# Clone the core library
git clone https://github.com/sciknow-io/skillful-alhazen.git
cd skillful-alhazen
# Clone example skills
git clone https://github.com/sciknow-io/alhazen-skill-examples.git