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/darpa_guidelines.md

View on GitHub
# DARPA proposal guidance

Reviewed 2026-10-01. DARPA funds high-risk research with national-security value,
including **basic and applied research**. Do not exclude fundamental work simply
because it is basic science: the [BAA guidance](https://www.darpa.mil/research/opportunities/baa)
explicitly covers both. The solicitation's technical scope and award instrument decide fit.

## Find and identify the opportunity

Start from [DARPA research opportunities](https://www.darpa.mil/work-with-us/opportunities)
and the official [SAM.gov](https://sam.gov/) record. Retrieve the solicitation,
amendments, attachments, Q&A and any proposer-day material. Record identifier,
revision, status, submission stages, deadline/time zone, authorized contact,
eligibility, award instrument, and submission channel. A program webpage or RFI is
not an open solicitation; a special notice requesting information is not an award offer.

The technical offices include BTO, DSO, IPTO, MXO, STO and TTO.
[In May 2026](https://www.darpa.mil/news/2026/what-is-in-name-darpa-reflecting-enduring-mission-future-focus),
I2O became **Information Processing Techniques Office (IPTO)** and MTO became
**Multi X Office (MXO)**. Older active records may retain earlier labels. Follow the
opportunity identifier and current amendments rather than rejecting them by name.

Program-specific BAAs define objectives; office-wide opportunities can solicit ideas
outside named programs. Some current opportunities use an Innovative Solutions
Opening or another mechanism, so do not force every application into an old BAA template.

## Establish fit before building the proposal

Use the [Heilmeier Catechism](https://www.darpa.mil/about/heilmeier-catechism) to organize
the investigator's reasoning. In your own words, answer:

1. What concrete result is sought?
2. What is the present baseline and limiting barrier?
3. What new idea makes progress plausible?
4. Who needs the result and what would it enable?
5. What are the main uncertainties and risks?
6. What resources and cost are needed?
7. What schedule is realistic?
8. What interim and final evidence will establish success?

Set demanding, justified metrics. Neither a 10x improvement nor a 50% failure
probability is a universal DARPA eligibility rule. Distinguish goals from preliminary
measurements; give baselines, test conditions and evaluation methods.

Program-manager discussion and proposer days can clarify fit when the solicitation
allows them. They are not universal mandatory prerequisites. Observe the designated
Q&A/contact channel and deadlines; do not contact evaluators outside that process.
Draft outreach for the user when asked, but do not send it without authorization.

## Map the required volumes

Use the actual attachment templates and volume/page limits. The following is an
outline to adapt, not a claim that every DARPA proposal has 20-40 pages or three phases.

| Element | Required substance |
| --- | --- |
| Summary/goals | Problem, measurable result, novelty, relevance and proposed effort. |
| Technical approach | Tasks, scientific rationale, preliminary evidence, assumptions, validation and alternatives. |
| Work statement | Deliverable work, task boundaries, responsible organizations and dependencies. |
| Milestones | Testable acceptance measures, dates, go/no-go decisions and resource consequences. |
| Team/resources | Relevant expertise and documented availability of facilities/data/materials. |
| Management | Coordination, risk tracking, reporting, decision authority and interfaces. |
| Transition | Intended user, operational constraints, remaining development and credible partner engagement. |
| Cost volume | Costs by required period/task/phase, basis of estimate, rates, subrecipient support and justifications. |

If phases are prescribed, show technical entry/exit criteria and avoid assuming later
options are guaranteed. If a government test team will evaluate outcomes, budget and
schedule the interfaces, test articles, documentation, and data handoffs. Do not claim
a transition commitment without evidence from the proposed partner.

## Budget and award instrument

DARPA can use procurement contracts, grants, cooperative agreements, and other
transactions. Select the allowed instrument and follow its cost, eligibility,
certification, data-rights and reporting provisions. Do not assume university status
guarantees a grant or that an other transaction automatically removes all conditions.

For each cost: connect the item to work, state quantity/rate/effort, identify the source
of the estimate, and reconcile totals across volumes. Separate fee/profit where
permitted; do not add it to assistance awards by analogy to contracts. Use authorized
institutional rates and properly obtained subcontract/subaward estimates. Cost realism
is not simply choosing the lowest request.

For SBIR/STTR, use the current topic and program instructions for small-business
eligibility, principal investigator employment, phase budgets, duration, and available
submission rounds. No historical Phase I/II amounts in this skill establish a current
cap. Young Faculty Award eligibility, award amounts and Director's Fellowship options
are solicitation-specific; a Director's Fellowship is not a generic independent call.

## Review and submission

Copy evaluation factors and their order/weight from the solicitation. Technical merit,
mission relevance and cost realism commonly matter, but there is no agency-wide
five-factor equal-weight rubric. Separate evaluation factors from selection/portfolio
considerations and availability of funds.

Use the exact submission portal or other channel authorized in the solicitation.
SAM.gov hosts notices; its listing alone does not mean uploading a proposal there is
the required route. Check registrations, institutional authority, file type/size,
volume separation, markings, signatures and deadline time zone. Public browsing does
not verify authenticated upload, receipt or acceptance. No submission API contract is
provided here.

Keep proprietary data and any controlled information in the authorized handling path.
For unclassified academic work, verify publication, intellectual-property and data
rights against the actual instrument and terms with the institution. Never place
restricted proposal content into an external image or language-model service merely
because a diagram would help.

## Final scientific checks

- The proposed advance is distinguished from existing results and competitors.
- Every quantitative metric has a baseline, measurement method and success criterion.
- A technically unsuccessful result still has a defined interpretation or fallback.
- Tasks, milestones, costs and team roles agree across volumes.
- Transition claims identify evidence, assumptions, and work remaining after the award.
- Reporting frequency, reviews and deliverables follow the solicitation/award rather
  than invented universal monthly or quarterly requirements.
- The institution has approved the assembled package and an authorized person
  confirms successful receipt; a draft or upload alone is not submission.
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