Tumbler — Methodology compiler

Assessment settings

—Indexed documents

—Branch heads in last index

—Indexed source hosts

Not runLatest index status

Loading corpus status…

Source coverage and exclusions

The registry contains hostnames discovered in Cloudflare plus the original WordPress wiki. The indexer is designed for current Git branch heads and published public HTML/text. Private API services and PDF document bodies are not included by default. Pages that fail or exceed crawl limits are reported rather than silently counted as complete.

Open corpus-indexing workflow ↗

Methodology / framework

Test assumptions, seek counterevidence and turn open questions into verifiable repository improvements.

What should the agent improve?

This adds direction to the work order. Scope defaults to all live Git branches; private service content is not silently ingested.

Property-law filterPrioritise title, parcel identity, conveyances, possession, occupation, registration, boundaries, court decisions and the legal regime applicable at the relevant date.

Fraud / concealment / government-involvement thresholdReview three separate hypotheses and require supporting evidence, counterevidence, independent corroboration and alternative explanations.

Evidential thresholdBalanced

ExploratoryCorroborated

Balanced: require multiple materially independent indicators and consider plausible alternatives. A missing record or refusal alone is insufficient.

Loading source inventory…

Generated agent prompt

Coverage is a factual status, not a promise. Tumbler uses retrieval over the latest index plus selected live source excerpts. It does not fine-tune a model. Until the full index has run and its coverage is checked, prompts must say so. Private API services and PDF bodies are excluded by default.

Choose a methodology and filters, then generate an implementation-ready work order for the agent.