Twenty-One Skills Make Award Intelligence Work Inspectable

The Labor as Code Skill Catalog

Companion to Surviving Scientific Review · Updated 2026-09-30

This is a starting catalog for an Award Intelligence AI Workforce. Each skill should have explicit inputs, allowed data sources, procedure, evidence requirements, output schema, prohibited actions, and audit criteria. The names are less important than making the work inspectable and repeatable.

“AI Lowers the Cost of Applying and Raises NIH’s Review Burden” explains the method. “Aristeion Is One Possible Software Version of This Method” shows one way to assemble the skills into a product.

Skill	What it does
Scientific Proposition Extractor	Turns a concept memo or Specific Aims draft into claims, evidence, uncertainties, dependencies, and PI-confirmation questions.
NIH Mechanism Mapper	Compares project maturity, scope, and investigator context with plausible NIH mechanisms; cites current source material.
NOFO Matcher	Finds relevant active opportunities and explains fit, exclusions, dates, and special review criteria.
Institute / Center Mapper	Uses mission statements, portfolios, RePORTER, and program pages to identify plausible funding homes.
Program Officer Mapper	Finds public NIH contacts associated with relevant programs and funded work and prepares questions, not predictions.
Study Section Mapper	Compares scientific content with current CSR scopes, shared interests, ART results, and historical funded-project review homes.
Reviewer Expertise Mapper	Summarizes expertise patterns from public rosters across multiple meetings without targeting named reviewers.
Funded Analog Finder	Retrieves and ranks close funded projects with evidence for why each analog is relevant.
Novelty Challenger	Tests claims of novelty against funded projects, publications, trials, and other public evidence.
Evidence Gap Auditor	Maps each major claim to supporting evidence and highlights leaps that may create avoidable reviewer skepticism.
Scope and Over-Ambition Auditor	Tests whether the amount of work, dependencies, and contingencies fit the mechanism and project period.
Budget and Scope Checker	Computes total costs from the direct-cost lines, each subaward site’s own rate, and the institution’s current rate; checks the modular line, the direct-cost threshold with consortium F&A excluded, and any total-cost cap in the NOFO; and lists what the sponsored programs office has to confirm.
Leadership Plan Red Team	Reads a Multiple PD/PI leadership plan, the effort table, and the site list the way the Investigator and Environment judgment will, and lists every question the plan leaves unanswered without rewriting it.
Review Criteria Auditor	Retrieves current mechanism/NOFO criteria and verifies that the application addresses each applicable dimension.
Assignment Request Builder	Prepares evidence-backed study-section, Institute, expertise-keyword, and conflict suggestions for PI review.
Program Officer Meeting Prep	Produces a concise concept brief, public context, questions, and decision points for a PO conversation.
Mock Review Orchestrator	Runs structured internal review roles, captures concerns, and maps them to criteria without pretending to predict NIH scores.
Compliance Package Auditor	Checks current application instructions, required attachments, formatting, and administrative conditions using authoritative sources.
Summary Statement Interpreter	Classifies critiques into science, evidence, clarity, fit, and preference or disagreement, flags administrative comments separately, and sorts each one by convergence, adoption, and criterion linkage to identify the likely score-driving issues.
Resubmission Architect	Helps the team decide what changes in an A1, maps every score-driving concern to a change and a location for the introduction, and preserves the distinction between rebuttal and improved science.
Outcome Reconciler	Compares pre-submission predictions with actual assignment, critiques, score, and award outcome to improve institutional memory.