Prompt library

These are the reusable prompt templates that drive the reliable-production method. They are the literal prompts used on BDPD, distilled into canonical, parameterised form and offered for reuse and scrutiny.

Prompt	Stage	Purpose
Artefact audit	Audit	The evidence-first review of one deliverable (paper, lecture, blog) against code and data, run by an independent reviewer.
Reviewer roles	Audit	Who runs the passes, the strengths each model brings, the failure modes each tends to exhibit, and the false-positive discipline.
Panel session	Ideate	Kickoff for a multi-model panel: convene several models around one question, converge on a shared deliverable.

How they fit together

A full verification of an artefact runs the artefact audit prompt once per reviewer (see reviewer roles), consolidates the reports, applies fixes, and repeats until convergence — the loop described in audit to convergence. The panel session prompt belongs to the earlier ideate stage, where a question is still open.

Each per-deliverable verification playbook also ends with its own copy-pasteable prompt specialised to that artefact; the templates here are the generic spine those specialisations share.

Conventions

• <artifact>, <reviewer>, <round>, <alias> are placeholders to fill before use.
• Paths are relative to the repository working tree.
• The golden rules in the method page apply to every prompt here; the audit prompt restates the load-bearing ones inline as an iron rule.