Blackmon Lab

The Scientist's Guide to Prompting

Most people treat AI like a search engine. This is a mistake.

To get value out of these tools, you need to understand that prompting is a hierarchy. It ranges from simple questions to massive, multi-page workflows that simulate an expert employee. Below, we break down the levels of prompting, culminating in the "Mega Prompts" used in our lab for manuscript review and grant writing.

---

1. The Hierarchy of Complexity

2. The "Mega-Prompt" Library

This is a repository of "Algorithmic Workflows"—complex, multi-stage instructions designed to force AI models to simulate expert roles.

How to use: Click the link to download the prompt text. Paste the entire block into a large-context model chat (like the top line models from ChatGPT, Claude, or Gemini). For best results, follow the instructions carefully most will require that you also upload documents to the chat. In models where you can select the level of thinking or speed always choose the "thinking" version and if possible use the paid version (typically you will be able to make use of more compute resources with paid versions). Access to some of these is provided by TAMU

Research & Writing Workflows (The "Scientific Method" Engines)

• The Manuscript Audit: Acts as an expert editor. Reviews your paper and helps you understand what needs to change. Think of it as a kind peer reviewer.
• The "Response to Reviewers" Diplomat: Takes raw/angry author comments and Reviewer 2’s critique, then outputs a table with "The Sandwich Method" (Agree, Refute w/ Evidence, Fix).
• The "Lay Summary" Translator: Converts dense genomic abstracts into 8th-grade reading level press releases, specifically flagging jargon like "heterozygosity."
• The "DDIG/Small Grant" Architect: Builds the narrative arc for a Dissertation Improvement Grant, ensuring the budget justification aligns perfectly with experiments.
• The "Literature Gap" Analyzer: Takes 5-10 abstracts as input and identifies the specific "knowledge conflict" that allows you to frame your research novelty.

Code & Data Science (The "Reproducibility" Engines)

• The "Coding Assistant: A prompt I use to fine tune the behavior of LLMs when I am using them to code. Pushes for certain appraoches (base R) and encourages the LLM to look for errors that are complex and not assume that I am a novice.
• The "R Package" Standardizer: Refactors a loose folder of R scripts into a formal R package structure (Description, Namespace, Roxygen2 docs).
• The "Bioinformatics Pipeline" Auditor: Reviews Bash/Nextflow pipelines for reproducibility gaps (missing versions, hard-coded paths).
• The "ggplot2" Publication Polish: Enforces Evolution or Systematic Biology formatting standards (font sizes, colorblind palettes, axis limits).
• The "SQL/Database" Schema Builder: Designs a normalized relational database schema for field collection data to prevent data rot.
• The "Phylogenetic Methods" Critic: Checks Methods sections for missing parameters in Bayesian or Likelihood analyses (priors, burn-in, bootstrap replicates).

Teaching & Mentoring (The "Pedagogy" Engines)

• The "CUREs Class" Syllabus Generator: A 14-week backward-design workflow that aligns weekly wet-lab deliverables with Bloom's Taxonomy learning objectives.
• The "Rubric" Constructor: Generates a fair, point-based grading rubric for subjective assignments (like posters) to standardize TA grading.
• The "Socratic Tutor" Generator: Consumes a topic (e.g., "Coalescent Theory") and generates a script for the AI to quiz the student iteratively instead of giving answers.
• The "Letter of Rec" Drafter: Drafts non-generic letters highlighting specific traits (resilience, troubleshooting) for Med School vs. Grad School based on your notes.

Graduate Student Survival (The "Stress Test" Engines)

• The "Quals Committee" Simulator Act as my PhD committee with knowledge of backgrounds and simulate my committee meeting with me.
• The "IACUC/IRB" Protocol Writer: Drafts "Search for Alternatives" and "Justification of Animal Numbers" sections based on a brief experiment description.
• The "Thesis Formatting" Automator: Checks a document against the specific, annoying formatting guidelines of the TAMU Office of Graduate Studies.
• The "Poster" Narrative Streamliner: Cuts text-heavy conference posters down to bullet points and visual cues, focusing on the "3-second hook."

Career & Administration (The "Bureaucracy" Engines)

• The "CV to Industry Resume" Translator: Translates academic "Dissertation Research" into industry-speak "Project Management" and "Data Analytics" for biotech jobs.
• The "Talk" Heckler: Generates the 5 most likely "Gotcha" questions a senior faculty member would ask during a tenure-track job interview.
• The "Conflict Resolution" Emailer: Drafts professional, boundary-setting emails to decline committee requests or address authorship disputes without burning bridges.
• The "Cover Letter" Tailor: Analyzes a specific journal's "Aims and Scope" and rewrites the "Why this journal?" paragraph to match their buzzwords.

Lab Management (The "Safety" Engines)

• The "SOP" Generator: Converts messy lab notebook scribbles into a formal, step-by-step Standard Operating Procedure with safety warnings.
• The "Equipment Grant" Justification: Writes the "Need" section for instrumentation grants, focusing on broader departmental impact.
• The "Safety Audit" Checklist: Generates a custom lab safety inspection checklist based on the specific chemicals and organisms listed in your inventory.