AI4RA Prompt Library¶
Versioned prompts, skills, agents, schemas, and component contracts for research-administration work.
This repository is the prompt-library leg of the AI4RA evaluation triad. The evaluation harness discovers components here, pairs them with datasets from AI4RA/evaluation-data-sets, and uses the shared ui-insight/AI4RA-UDM foundation where a component is UDM-aligned. Contract scope is recorded per component in component_catalog.json.
Start by filtering below, or jump to:
- Browse by category · by domain · by status · by tag
- Contracts · About · Taxonomy · Contributing
- Source:
AI4RA/prompt-library· Machine discovery:component_catalog.json
Browse all components¶
award-compliance-extraction-udm
experimentalExtracts the compliance framework and financial-management requirements from a federal award document into a single structured JSON object that drives a consolidated "Award Compliance & Financial Overview" deliverable for post-award setup and monitoring. Produces two main blocks (compliance framework with 10 fields including a normalized compliance calendar; financial management with 10 fields including budget periods and budget categories).
document-type-classifier-udm
experimentalClassifies a research-administration document into a controlled type vocabulary so downstream pipelines can route it to the correct extractor or reviewer. One call in, one small JSON object out.
effort-reporting-extraction-udm
experimentalExtracts effort-reporting and personnel-compliance requirements from a federal award document into a structured JSON object that drives an "Effort Compliance Brief" for post-award compliance tracking. Captures the reporting cadence, certification method (per 2 CFR 200.430(i)), PI commitments, per-person key-personnel commitments, cost-shared effort, governing regulation, record-retention requirement, and referenced governing documents.
ffr-management-extraction-udm
experimentalExtracts Federal Financial Report (FFR / SF-425) requirements and post-award financial reporting obligations from a federal award notice or agreement into a structured JSON object organized around the five buckets a sponsored-programs analyst uses when preparing FFR submissions: submission schedule, submission system and procedures, required financial data, compliance consequences, and preparation timeline.
foa-checklist-extraction-udm
experimentalExtracts a Federal Funding Opportunity Announcement (FOA) into a structured JSON object covering the eight reference sections a federal-grants office uses for FOA review: FOA summary, key dates, funding information, eligibility, application components, evaluation process, program priorities, and special requirements. Stronger emphasis than its sibling `rfa-checklist-extraction-udm` on the **evaluation criteria**, **review process**, and **submission system** fields.
nsf-award-notice-extraction-udm
stableExtracts an NSF Award Notice (initial obligation or amendment) into a single JSON object conforming to this repo's UDM-aligned extraction contract. Designed for the NSF notice format that arrives as a PDF printed from Outlook after OSP receipt — the email header, boxed notice body, and NSF-format 18-category budget table.
nsf-budget-justification-render-udm
experimentalRenders a reviewed eight-section NSF budget-justification array into Word-pasteable Markdown and HTML strings. The component is the formatting tail of a multi-step budget-justification pipeline: it does not draft, review, or reinterpret content — it transforms the structured array into two presentation forms that paste cleanly into a Microsoft Word document.
nsf-budget-justification-review-udm
experimentalValidates and polishes a drafted NSF budget-justification array against its source structured budget and NSF narrative conventions. The component is the QA step of the multi-step budget-justification pipeline: it checks section completeness and ordering, fixes misplaced content (for example graduate tuition in the wrong section), ensures required NSF disclosures are present, and consolidates terminology across sections. Its revised array is the input to the downstream render step that emits Word-pasteable Markdown and HTML.
nsf-budget-justification-udm
experimentalDrafts an NSF-format budget justification narrative from a structured budget object. Output is the eight canonical NSF sections (A..H) as an ordered array of section objects — one per section, each with a `key`, a fixed `title`, and `content` in narrative markdown. All eight sections are always emitted; zero-value categories produce a single-sentence content.
nsf-budget-spreadsheet-ingest-udm
experimentalNormalizes an NSF-style proposal budget workbook (or workbook-derived evidence) into the structured budget object consumed by `nsf-budget-justification-udm`. This component is the extraction half of a multi-step budget-justification pipeline: it isolates spreadsheet interpretation from narrative drafting so each responsibility can be prompted, versioned, and evaluated independently.
nsf-budget-spreadsheet-justification-udm
experimentalDrafts an NSF-format budget justification directly from an NSF-style proposal budget spreadsheet or spreadsheet evidence package. The component interprets workbook tabs for personnel, equipment, travel, participant support, other direct costs, subawards, tuition/fees, and rates, then emits the same eight-section JSON narrative contract used by `nsf-budget-justification-udm`.
nsf-expense-allowability-check
experimentalReviews a single expense against NSF award terms, supplied budget evidence, and general sponsored-programs allowability criteria. The component turns the original checklist-style "Expense-Allowability Check" prompt into an evidence-grounded chat review with explicit rule checks, budget alignment, citations, follow-up actions, and a conservative final decision.
prior-approval-extraction-udm
experimentalExtracts all prior-approval requirements from a federal award document into a structured JSON object suitable for driving a tracking system of activities that need federal-agency authorization before they can be executed. Produces three logical buckets (budget-related, scope/timeline, approval procedures table) plus a list of Research Terms and Conditions (RTC) waivers and expanded-authority delegations.
proposal-budget-personnel-extraction-udm
experimentalExtracts personnel information and compliance-requirement triggers from a proposal budget document into a structured JSON object. Identifies senior key personnel, postdocs, graduate students (RA / TA), undergraduates, and other personnel; the budget category structure; subaward recipients, equipment over $5,000, travel summary, F&A rate and base, cost sharing, and total costs. The output drives the downstream `proposal-document-completeness-udm` gap analysis.
proposal-document-completeness-udm
experimentalAutomates the near-final document-completeness review of a proposal package: identifies all senior key personnel, the four required documents per person (biosketch, current & pending, collaboration & affiliation, synergistic activities), subaward documents, and conditional-requirement triggers. Produces a single gap-analysis JSON for sponsored-programs analysts to use when sending a "missing documents" message back to the PI.
rfa-checklist-extraction-udm
experimentalExtracts a federal funding announcement (RFA / FOA / NOFO / program solicitation) into a structured JSON object organized around the **eight-section pre-award checklist** that a sponsored-programs analyst actually uses when triaging a new opportunity. The shape enforces placement rules — award amount lives in `award_information` only, detailed financial rules in `budget_requirements` only — so downstream consolidation does not have to re-adjudicate where a fact belongs.
rfp-extraction
stableExtracts a comprehensive submission and eligibility checklist from any federal funding announcement — RFP, RFA, FOA, NOFO, BAA, Dear Colleague Letter, or program solicitation. Output is a single markdown document with GFM checkboxes that PIs can use for proposal development and OSP staff can use for pre-submission compliance review.
rfp-extraction-udm
stableExtracts any federal funding announcement (RFP, RFA, FOA, NOFO, BAA, Dear Colleague Letter, or program solicitation) into a single JSON object conforming to this repo's UDM-aligned extraction contract. Companion to [`rfp-extraction`](../rfp-extraction/), which produces the human-readable checklist form from the same source.
solicitation-doc-modifications-udm
experimentalGiven a solicitation text and a sponsor's default document requirements, emit only the **diff**: modifications to the defaults and net-new documents the solicitation introduces. Each entry links back to the default it overrides (when applicable) via `modifies_default` and carries a verbatim `source_excerpt` from the solicitation.
sponsor-doc-defaults-udm
experimentalGiven a sponsor organization (and optional division), emit the default set of documents that sponsor typically requires for a standard full proposal — as a structured list with page limits, format specs, required vs. optional flags, per-person indicators, and conditional triggers.
subaward-extraction-udm
experimentalExtracts a fully executed subaward agreement (Pass-Through Entity → Subrecipient) into a single structured JSON object covering nine sections used for research-administration setup and ongoing monitoring: basic info, project periods, PTE contacts, subrecipient contacts, financial summary, financial policies, reporting requirements, prior-approval handling, and key compliance requirements.
vandalizer-to-udm-translation
experimentalConverts a Vandalizer NSF-extraction JSON object (flat key/value form) into a single JSON object conforming to the `nsf-award-notice-extraction-udm` schema. This is a pure transformation — no information is invented, and fields Vandalizer does not capture emit as `null` or documented defaults.