Ontolog

Ontolog is an open-source agent skill and knowledge repository starter for documenting how a business works. Contributors explain their terminology, rules and exceptions through interviews. Their perspectives become attributed Markdown records with YAML metadata, examples, unresolved questions and scoped approval decisions. The build combined interview instructions, a portable repository template and Python tooling for initialization, skill installation, ZIP export and validation. Agents helped implement the tooling, exercise synthetic contributor interviews, review changes independently and configure GitHub security checks. Synthetic evaluations preserved differing definitions and supported interrupted interviews. They also exposed a consistency problem: recording an approval could leave a related question incorrectly unresolved. Repeated fresh sessions informed an instruction change. Later fixes tightened approval-period validation and converted excessive YAML nesting from a traceback into a normal validation error. This remains an alpha. Synthetic interviews do not establish effectiveness with real organizations, and native-app imports remain unverified. Project-wide usage totals could not be reliably isolated from shared-directory and mixed-project sessions without inspecting conversation text.