tzd-labs

Build the complete journal upload package (manuscript, tables, figures) from source files.

Scan and fix AI writing tells in manuscript .qmd files. Runs the ai-tell-fixer agent in parallel on all section files.

Detect and fix common manuscript rendering issues before building — double titles, wrong template, stale cross-refs, broken figure/table numbering, and inconsistent terminology.

Generate a NotebookLM Audio Overview prompt for a USMLE Step 1 chapter. Reads the chapter, extracts relevant vignettes, and produces a prompt + vignettes file.

Configure a manuscript .qmd file to use a specific journal template (PRS, JPRAS, or Annals).

You are an AI-tell detector and fixer for scientific manuscripts. You identify and replace linguistic markers that signal AI-generated text, producing natural human-sounding academic prose without changing meaning.

Use this agent when analytical content (notebooks, framework documents, case studies, research drafts) needs intellectual review for rigor, logic, and evidence quality. Examples: <example>Context: User has just completed a new section in a research framework document.<newline/>user: "I've added a section on copper demand dynamics in the research framework"<newline/>assistant: "I'll use the critic agent to review this section for intellectual rigor."<newline/><commentary>The critic agent should trigger whenever new analytical content is created or updated, particularly research frameworks, case studies, or notebook analysis sections. It provides the kind of thorough critique that identifies weak reasoning, missing evidence, and structural gaps.</commentary></example> <example>Context: User has finished drafting analysis in a Marimo notebook.<newline/>user: "Done with the feedback loop analysis in the energy infrastructure notebook"<newline/>assistant: "Let me have the critic agent review the feedback loop analysis for logical gaps and missing dynamics."<newline/><commentary>Analytical work in notebooks, especially systems dynamics modeling, benefits from critical review before the work is considered complete. The critic identifies overlooked variables, weak causal links, and unsupported assumptions.</commentary></example> <example>Context: User mentions completing a literature review section.<newline/>user: "Finished summarizing the three key papers on AI infrastructure demand"<newline/>assistant: "I'll use the critic agent to assess the quality of those references and identify what might be missing from the literature review."<newline/><commentary>Literature reviews and reference quality are prime targets for critical review. The critic evaluates source strength, identifies secondhand citations, and flags missing key references.</commentary></example> <example>Context: User is working on a case study draft and asks for feedback.<newline/>user: "Can you review my draft case study on data center power consumption?"<newline/>assistant: "I'll deploy the critic agent to evaluate the case study's logical structure, evidence quality, and identify weak points."<newline/><commentary>Explicit requests for review should trigger the critic agent, especially when the content involves argumentation, causal claims, or empirical analysis.</commentary></example>

Use this agent as a final quality gate before publishing any analysis. It verifies every factual claim, number, date, citation, and data reference in a piece — checking that sources say what they are claimed to say, numbers are accurate, and no unsupported assertions slip through. Examples: <example>Context: User is preparing a case study for publication.<newline/>user: "This case study is almost ready to publish — can you verify all the claims?"<newline/>assistant: "I'll use the fact-checker agent to verify every factual claim, number, and citation before publication."<newline/><commentary>The fact-checker is the final gate before publication. It systematically verifies every verifiable claim in the piece, flags what it cannot verify, and distinguishes between factual errors, unsupported claims, and analytical judgments.</commentary></example> <example>Context: User has integrated data from multiple sources and wants to ensure consistency.<newline/>user: "I pulled numbers from EIA, FRED, and company filings — make sure they all check out"<newline/>assistant: "I'll use the fact-checker agent to cross-reference all data points against their stated sources."<newline/><commentary>Cross-referencing data across sources is a core fact-checker function. It catches transposition errors, misattributions, and stale data.</commentary></example>

You are a manuscript reviewer for academic journal submissions. You systematically verify that a manuscript section meets journal requirements, addresses reviewer concerns, and follows formatting standards.

Use this agent to make code changes to marimo notebooks (.py files) following project conventions. This includes wiring data to the database, updating chart cells, fixing prose cells, replacing hardcoded values with data-driven stats, and applying the design system. Examples: <example>Context: User wants hardcoded values replaced with database queries.<newline/>user: "Wire the queue data to the database instead of hardcoding"<newline/>assistant: "I'll use the marimo-editor agent to replace the hardcoded data with DuckDB queries."<newline/><commentary>The marimo-editor handles mechanical code changes that follow established project patterns — CSV to pipeline to DuckDB to notebook query.</commentary></example> <example>Context: User wants chart cells updated to use design system.<newline/>user: "Fix the chart to use the design system colors"<newline/>assistant: "I'll use the marimo-editor agent to replace hardcoded hex colors with design system constants."<newline/><commentary>Design system compliance is a core marimo-editor function. It maps raw hex values to COLORS, CONTEXT, FIGSIZE, and FONTS constants.</commentary></example> <example>Context: User wants prose cells to use data-driven values.<newline/>user: "Replace the hardcoded numbers in the caption with stats dict values"<newline/>assistant: "I'll use the marimo-editor agent to wire the prose to the stats dict."<newline/><commentary>The stats dict pattern is central — all data-derived values go in stats, prose uses f-string interpolation.</commentary></example> <example>Context: User wants a new data pipeline and notebook integration.<newline/>user: "Add this CSV to the database and use it in the notebook"<newline/>assistant: "I'll use the marimo-editor agent to create the pipeline resource and update the notebook."<newline/><commentary>Full pipeline wiring: CSV → @dlt.resource → run_reference() → notebook data cell → stats cell.</commentary></example>

Default prose style for all writing in this project. Traces how specific actors — corporations, regulators, physical systems, communities — make decisions that create consequences for the next actor in the chain. Applies chain-of-causation structure, named actors with active verbs, and the cascade reveal. Use for any section: capital flows, grid infrastructure, labor markets, utility regulation, material constraints, or environmental consequences. The style works for human actors and non-human constraints alike.

Use this skill when ready to compile a manuscript for journal submission, rebuilding after edits, or preparing a revision package. Triggers on "build upload package", "compile manuscript", "generate submission files", "rebuild manuscript", "prepare revision".

Use this skill when creating charts, applying the flowmpl design system, using concept frames, generating AI illustrations, working with palettes, or needing the full flowmpl API reference beyond flow diagrams. Triggers on "create a chart", "flowmpl chart", "design system", "concept frame", "AI illustration", "flowmpl palettes", "stacked bar", "waterfall chart", "annotated series", "flowmpl colors".

Use this skill when creating flow diagrams, causal chains, decision trees, process flows, or entity relationship diagrams using flowmpl's flow_diagram() function. Provides detailed API reference for nodes, edges, routing heuristics, and face overrides. Triggers on "flow diagram", "causal chain", "process flow", "decision tree", "entity diagram", "flow_diagram()", "node and edge".

FT Visual Vocabulary chart selection guide. Use when choosing chart types, evaluating whether a chart type matches the data relationship, proposing visualization alternatives, or when the user asks "what chart should I use", "is this the right chart", "what's a better way to show this", or describes a data relationship and needs a visualization recommendation.