Skip to content

K-Dense-AI/scientific-agent-skills / skills/research-grants

research-grants

Verified

Supports research proposal preparation and review for NSF, NIH, DOE, DARPA, and Taiwan NSTC, including opportunity-specific requirements, aims, review criteria, budgets, broader impacts, forms, and resubmissions. Use for investigator-authored grant development, compliance matrices, and proposal critiques.

references/nih_guidelines.md

View on GitHub
# NIH application guidance

Reviewed 2026-10-01. Start from a specific active NOFO, its amendments, related NIH
Guide notices, due date, activity code, and application form package. Since October
2025, [Grants.gov is the official source of NIH funding opportunities](https://www.grants.nih.gov/funding/explore-nih-opportunities);
the NIH Guide continues to carry policy notices. An old Guide NOFO or a RePORTER award
is not proof of current availability or eligibility.

## Appropriate assistance and applicant ownership

[NIH policy](https://grants.nih.gov/grants/policy/nihgps/HTML5/section_2/2.3.7_policies_affecting_applications.htm)
permits limited AI assistance but will not consider applications, or sections,
substantially developed by AI to be applicants' original ideas. Use this skill to
extract requirements, critique investigator-authored material, check consistency and
calculations, and suggest limited edits. Keep the scientific ideas, interpretation,
and substantive application development with the applicant; human review alone does
not erase the policy. Never invent preliminary findings, citations, collaborators,
patient access, credentials, or commitments. The same policy section limits an
individual PD/PI or MPI to six covered applications per council calendar year, with
specified activity-code exclusions; verify how the limit applies before planning
multiple submissions.

## Review framework

For most research project grants with due dates on/after January 25, 2025, use the
[simplified review framework](https://www.grants.nih.gov/policy-and-compliance/policy-topics/peer-review/simplifying-review/framework):

| Factor | Applicant's evidence | Assessment |
| --- | --- | --- |
| Importance of the Research | Important gap and contribution; significance and innovation | 1-9 |
| Rigor and Feasibility | Design, analysis, sample-size justification, bias control, alternatives, timeline | 1-9 |
| Expertise and Resources | Project-specific capabilities, facilities, access and missing expertise | Sufficiency, not a separate numeric score |

All three affect overall impact. This is not a weighted average. Additional review
criteria, including applicable human-subject and animal protections, may affect
impact despite lacking standalone scores; budget appropriateness is a different
review consideration. Check Section V of the NOFO: fellowships, K awards, small
business, and complex multi-project mechanisms need their own review instructions.
The research narrative still uses Significance, Innovation, and Approach where the
application instructions require them; do not rename these to the review factors.

[NOT-OD-26-114](https://grants.nih.gov/grants/guide/notice-files/NOT-OD-26-114.html)
continues modified practices for the January 2027 Advisory Council: most meetings
discuss 30-35% of applications; a middle category is competitive but not discussed.
Such applications can still be considered for funding. Summary statements are
simplified. Do not promise that half will be discussed or classify an application as
unfundable solely because it was not discussed. Overall impact scores for discussed
applications normally run 10-90; do not map a score to an agency-wide funding chance.

## Narrative and attachments

Use the [current page-limit table](https://grants.nih.gov/grants-process/write-application/how-to-apply-application-guide/page-limits)
and the NOFO. For standard R01 applications, Specific Aims is **up to one page** and
Research Strategy up to 12; for standard R03/R21 the strategy is generally six. The
Project Summary/Abstract is generally at most 30 lines, and Project Narrative at most
three sentences. These are maxima, not required exact lengths or interchangeable
components. Check exceptions rather than shrinking type to fit.

An aims page should connect the gap, objective, investigator's hypothesis or research
question, achievable aims, expected knowledge, and impact. Two or three aims may work
well, but NIH has no universal required number. Avoid making all later aims depend on
one uncertain result; state alternatives. See [specific_aims_guide.md](specific_aims_guide.md)
for illustrative writing patterns, not a ready-to-submit application.

For the research strategy:

1. **Significance:** identify the important barrier and what resolving it changes.
   Cite the strengths and limitations of prior evidence, not just disease burden.
2. **Innovation:** explain the actual conceptual or technical advance. Do not claim
   that all worthwhile NIH research must be revolutionary.
3. **Approach:** connect each aim to design, subjects/models, controls, randomization
   and blinding where appropriate, sample-size rationale, analysis, quality controls,
   plausible outcomes, pitfalls, alternatives, resources, and a realistic timeline.
   Separate observed preliminary results from predictions. Do not present a power
   calculation as meaningful without its effect-size/variance assumptions.

Follow [attachment requirements](https://www.grants.nih.gov/grants-process/write-application/how-to-apply-application-guide/format-attachments):
PDF; ordinarily at least 11-point body text, no more than 15 characters per horizontal
inch or six lines per vertical inch, and at least half-inch margins. Smaller graphic
text may be acceptable if legible at 100%. Check searchable text, embedded fonts,
filenames and file limits in the actual package. Do not put overflow narrative in
appendices or hyperlinks. Preserve SciENcv-generated certification in its PDF instead
of printing or altering it to imitate a manually formatted attachment.

## Common Forms, disclosures, and data sharing

For application due dates and relevant JIT/RPPR/prior-approval submissions on/after
January 25, 2026, NIH requires Common Forms; enforcement began May 8, 2026.
Use [SciENcv for the Biographical Sketch Common Form and NIH Supplement](https://www.grants.nih.gov/grants-process/write-application/forms-directory/biographical-sketch-common-form)
and the Current and Pending (Other) Support Common Form. There is no downloadable
blank substitute for the SciENcv biosketch. Each investigator must review and certify
accurate records, supply the required ORCID linkage to eRA Commons, and follow the
current field limits. Do not reuse the retired five-page Word biosketch format.
Other Support includes applicable external support and in-kind resources; reconcile
scientific, budgetary, and commitment overlap with the actual instructions.

For data-generating activities covered by the DMS policy, use the updated structured
plan format required for competing applications from May 25, 2026 under
[NOT-OD-26-046](https://grants.nih.gov/grants/guide/notice-files/NOT-OD-26-046.html).
[NOT-OD-26-100](https://grants.nih.gov/grants/guide/notice-files/NOT-OD-26-100.html)
explains JIT/RPPR transitions and, effective October 1, 2026, reporting plan changes in
RPPR C.5.c rather than obtaining DMS-plan prior approval. Check applicability and
format for the mechanism; do not substitute a generic NSF plan. Cost data management,
curation and sharing explicitly, with justified access restrictions.

## Mechanism selection

- **R01:** individual research project; not limited to established investigators.
  Preliminary studies normally support feasibility; the instructions and NOFO determine
  requirements. There is no universal $250,000 minimum or maximum award budget.
- **R21:** exploratory/developmental; the [activity-code guidance](https://www.grants.nih.gov/funding/activity-codes/r21)
  describes up to $275,000 direct costs across two years, no more than $200,000 in one
  year, and no required preliminary data. It is not renewable; that is different from
  a categorical ban on no-cost extensions. Check participating ICs and the NOFO.
- **R03:** small-scope research; use the active NOFO for budget, period and eligibility.
- **R15:** distinguish AREA and REAP eligibility; institutional research funding and
  student populations matter, not simply a label of "undergraduate institution."
- **K and fellowship awards:** check degree, citizenship/residency, career stage,
  protected effort, mentor/reference requirements, and IC-specific salary/research
  allowances. A K award is not a research-project grant with an extra training paragraph.
- **U/P/R35 and other codes:** use mechanism-specific scope and instructions. R35 is
  not synonymous with the NIGMS MIRA program across NIH.

[NOT-OD-26-098](https://grants.nih.gov/grants/guide/notice-files/NOT-OD-26-098.html)
announces K-program changes for due dates on/after **October 12, 2027**, including
consolidated codes, DMS-plan changes and clinical-trial restrictions. Do not apply
those future rules to an October 2026 application merely because the notice exists.

## Budget

Select the budget form using the NOFO and organization eligibility. For eligible
applications, [modular budgets](https://grants.nih.gov/grants/developing_budget.htm)
request $25,000 modules, up to $250,000 direct costs per budget period **excluding
consortium F&A**. Foreign institutions and SBIR/STTR do not use the modular format.
Build a detailed internal estimate even when only a modular narrative is submitted.
Modular personnel justification states names, roles and person-months, not salary or
fringe rates. Explain variation in modules; there is no generic $125,000/25% rule.

The FY2026 Executive Level II cap is $228,000 effective January 1, 2026, as corrected by
[NOT-OD-26-038](https://grants.nih.gov/grants/guide/notice-files/NOT-OD-26-038.html).
The September 28 [continuing-resolution notice](https://grants.nih.gov/grants/guide/notice-files/NOT-OD-26-131.html)
continues FY2026 fiscal policies through December 11, 2026. Recheck for later awards.
Use actual institutional salary, appointment basis and effort; the cap is not a salary
recommendation. There is no general one-month PI effort limit. An NRSA stipend is not
an employee postdoctoral salary scale.

Use the applicable negotiated F&A rate and base. **NIH does not impose a blanket 8%
F&A rate on domestic subawards.** Do not apply the revised federal $50,000 MTDC
subaward threshold automatically: [NOT-OD-26-072](https://grants.nih.gov/grants/guide/notice-files/NOT-OD-26-072.html)
rescinded NIH implementation of revised MTDC/de-minimis provisions. Verify the
applicable $25,000 treatment, agreement, and current fiscal instructions with the
institution. Equipment classification also uses the current NIH form instructions
and the institution's lower capitalization threshold where applicable.

Foreign collaborations no longer use ordinary nested foreign subawards for new
applications. Consult the [current subaward policy](https://www.grants.nih.gov/policy-and-compliance/policy-topics/subawards)
and [PF5 instructions](https://grants.nih.gov/funding/activity-codes/pf5) for eligible
linked domestic/foreign award routes. Do not infer eligibility from a legacy budget
webpage's mention of an 8% foreign consortium rate.

## Submission and deadlines

Use the NOFO's Key Dates and current
[standard-date table](https://www.grants.nih.gov/grants-process/submit/submission-policies/standard-due-dates).
For example, standard *new R01* dates are February/June/October 5; R01 A1 dates differ.
NIH no longer uses dedicated AIDS deadlines after the 2026 transition. Dates are
normally 5 p.m. applicant-organization local time; check weekends, federal holidays,
late-submission rules and any emergency notices. Do not infer eligibility for late or
continuous submission from old service history.

ASSIST, Grants.gov Workspace, and supported institutional system-to-system software
are preparation/submission options. ASSIST routes through Grants.gov; eRA Commons is
used to track receipt, errors and the assembled application, not an alternative
public submission API. Confirm UEI/SAM, Grants.gov, organization/PI Commons accounts,
affiliation and authorized signing roles well in advance. Do not assume registration
will finish within a fixed number of days. Review validation errors and the assembled
application within the applicable viewing/correction window.

[NOT-OD-26-094](https://grants.nih.gov/grants/guide/notice-files/NOT-OD-26-094.html)
announces replacement of letters of support with short letters of collaboration at
**FORMS-J implementation in FY2027**. It is not a blanket instruction to relabel
FORMS-I attachments now. Confirm the implementation date and assigned form package;
when effective, the notice limits each collaboration letter to 100 words and removes
endorsements. Career/training commitments and confidential reference letters have
distinct instructions.

## Resubmission and feedback

NIH permits one A1 within 37 months of the initial application due date when the NOFO
accepts resubmissions. Follow the [resubmission policy](https://grants.nih.gov/grants-process/submit/submission-policies/resubmission-applications),
including waiting for the summary statement and restrictions on overlapping review.
A subsequent new A0 is not an unrestricted right to simultaneous duplicate submission.

For a standard research A1, use an introduction of up to one page responding to major
criticisms, then implement the changes throughout the application. Use exact final
section/page references and explain disagreements with evidence. Do not mark changes
by bolding, brackets, shading or track changes solely to identify revisions. Recheck
limits for other application types. Decide what to revise from reviewer concerns and
program fit, not invented score bands or historical payline predictions.
add-product-e2e-evalpaperclipai/paperclipAdd or extend a Paperclip full-stack runner E2E workflow, fixture, matcher, or report evidence path for local or Daytona execution.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
add-runner-evalpaperclipai/paperclipAdd or extend a Paperclip Runner protocol evaluation definition, roster, assertion, or report fixture with provenance and narrow validation.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
diagnose-why-work-stoppedpaperclipai/paperclipCompatibility entry point for failed, stalled, looping, or over-recovered Paperclip work. Use fix-failed-paperclip-run for the full diagnosis, regression, repair, real verification, and green-PR workflow; honor an explicitly requested diagnosis-only scope.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
fix-failed-paperclip-runpaperclipai/paperclipDiagnose and fix failed, stuck, slow, or incorrectly blocked Paperclip tasks and runs. Use with a task/run URL or failure report to trace the root cause, prove a regression red then green, check second-order effects, verify the real workflow, and deliver a PR with passing checks.Stars: 115kUpdated Claude CodeCodexVerifiedInstall