Prompt Engineering Tools
Tools for working with AI, written as plain files. Read what each one is for, then download it as a Markdown (.md) or text (.txt) file and keep it on your own computer.
Hallucination Guard
A drop-in prompt that makes an AI stop inventing things. It forbids fabricated names, quotes, numbers, dates and citations; makes the model check a question’s premise before answering it; requires calibrated language (known, likely, uncertain, unknown) instead of confident guessing; and requires the arithmetic to be shown so it can be checked. It bans “compliance theater,” where a reply recites the rules without following them. Paste it at the start of a conversation or into a system prompt. Current version: 1.3.
Spiritual Discernment Tool (SDT)
Evaluates an article, transcript, sermon or piece of media at two levels. The first prompt (SDT-A) does a structural analysis of what the content claims, how it argues, and where it is solid or weak. The second (SDT-B) is a discernment layer that asks what the content forms in the person who takes it in: whether it produces the fruit of the Spirit or a counterfeit of it. Use it when you need more than a fact-check, when the question is not only “is this accurate?” but “what will this do to me?” Current version: 5.1.
Media Quality Grader (MQG)
Grades a news article, video or other media item for how trustworthy it is, and tells you why. It scores seven weighted dimensions: sourcing (25%), verification (15%), framing (15%), voice and attribution (15%), context (10%), transparency (10%), and ideological anatomy (10%), which extracts the “near-truth” a piece is built around. Ceilings stop one strong area from hiding a serious failure in another. The output is a letter grade, the specific findings behind it, and a plain Reader Takeaway for someone who will not read the full analysis. Current version: 12.5.
Theology Grader
Grades a sermon, essay or teaching on how well it carries the Christian message, and keeps the grades separate so one strength cannot hide a weakness. It scores theosis depth (30), love as God’s own nature in the Johannine sense (30), communicative effectiveness (30), and apostolic fullness (35). An orthodoxy floor and a set of falsification stages come first, so a piece that fails on basic doctrine cannot earn a high grade by being well written. Version 8.0 adds a verdict-lock sentence for lower-tier results so the verdict is stated plainly and not softened. Current version: 8.0.
Absolute Transparency Standard (ATS)
A framework for designing or auditing communication that comes in installments: a sermon series, a newsletter run, a campaign, a long document. It applies seven tests (Roadmap, Standalone, Borrowed Trust, Jump-Ahead, Reversibility, Stacking and Keystone) to ask whether a reader is told where the argument is going, can judge each piece on its own, and can walk back out if they disagree. It separates honest sequencing from drip-fed persuasion by one measure: whether the plan is disclosed. Use it in design mode while writing or in audit mode on something already published. Current version: 1.2.
Canonical Context Brief (CCB)
A preparation step to run before a sermon or teaching is drafted. It first asks whether you already have a message in mind or are letting the text tell you where to go, then walks the assigned passage through seven exegetical passes: the superscription, canonical cross-references, and the passage’s place in redemptive history, among others. The result is a brief of what the text itself establishes, so the draft starts from the passage and not from a point you brought to it. It does not replace your own study and it does not write the sermon. Current version: 1.6.