pubskill-lib · repository evidence
Choose the question first. Each surface states what it can and cannot prove.
Inspection
What obvious repository defects and unresolved static boundaries can be found without executing it?
Free static inspection. It does not install dependencies, run tests, execute repository code, or verify runtime behavior.
Examiner
What is actually in this codebase, how is it structured and documented, what can be measured safely, and where are the unresolved boundaries?
Structural examination does not require AI. Narration is optional.
Inspection
The inspector checks repository identity and selected static surfaces such as broken local links, obvious CI no-ops, missing declared Python entry points, and directly referenced local script targets.
A true repository audit is not offered yet. It requires isolated execution of applicable repository gates plus reproducible receipts. Static inspection remains useful, but it is not that.
Examiner
The Examiner inventories source, checks safe adapters and placement boundaries, measures supported RATIOS, detects existing NARRATIVE blocks, preserves unresolved conditions as hmmm, and assembles repository documentation.
Without AI
Inventory, measurable structure, safe write boundaries, RATIOS, existing narratives, documentation assembly, and unresolved boundaries.
Optional AI
Add evidence-bound source narration. The user keeps the choice between their own AI provider and hosted AI.
The Examiner CLI is available now. Hosted Examiner execution, pricing, and metering remain unresolved until measured from real repository runs.