grant-writing
Plan, draft, and revise competitive research grant applications — research plans, mentoring/career-development plans, budget justifications, abstracts, and specific aims. Invoke whenever writing or st
SKILL.mdskills/grant-writing/SKILL.md
GitHubGrant writing — competitive research & career-development awards
Goal: help build a proposal that is responsive to the funder, rigorous, and compelling — one that survives a fast screening read and rewards a deep one.
Workflow (in this order)
- Read
LEARNINGS.mdbeside this skill, when your copy keeps one, for funder-specific rules and what reviewers rewarded/penalized in prior rounds. - Audit responsiveness before touching prose — pull the funder's required-components list + review criteria and walk the draft element-by-element (method below).
- Fix gaps in priority order — focus-area alignment first, then down by impact, with surgical edits confined to the affected section.
- Draft/revise using the component-craft sections below; for voice, pair with a
scholarly-voice/academic-writing skill if you keep one (this install's:
LOCAL.mdbeside this skill, when your copy keeps one — Workflow step 4); defer to research-methods for study design and human subjects. - Final pass — run the "Before you call this done" checklist, then record what this round taught in LEARNINGS.md.
The first move is always responsiveness, not prose
Before drafting or polishing anything, audit the draft element-by-element against the funder's published required-components list and review criteria. A beautifully written proposal that is out of scope for the funder loses. Name the gaps by category first, then fix in priority order.
The single highest-leverage check: does the proposal center what the funder funds?
- WTG Scholars funds programs, policies, or practices that reduce inequality in youth outcomes (ages 5–25). It explicitly does not fund studies of the causes, extent, or consequences of inequality, or that merely describe it.
- The most common fatal flaw found in review: the plan centers the background problem
as a cause of disparities (out of scope) instead of the practice under study that
reduces a named disparity. The fix is to name the practice up front and recast the
rest as rationale. (This install's worked case:
LOCAL.mdbeside this skill, when your copy keeps one — Fatal-flaw worked case.)
Generalize: for any funder, identify the unit they fund (a practice / intervention / mechanism / discovery) and make sure that unit — not the background problem — is the subject of your Significance and aims.
Responsiveness-review method (reusable)
- Pull the funder's required-components list + review criteria verbatim.
- Walk the draft section by section; for each required element, mark present / weak / missing and why.
- Group gaps by category. The categories that recur in research-plan reviews:
- Focus-area alignment — centering a cause/consequence instead of a fundable practice (most critical).
- Outcome not specified — no named, literature-evidenced outcome inequality with its causal mechanism.
- No standalone Human Subjects Protection section — IRB / data security / confidentiality / risk / mitigation scattered instead of dedicated.
- Analysis plan underspecified — methods listed but no analysis/interpretation framework (a list of data sources is not an analysis plan).
- Generalizability & power not addressed — single-site transferability and statistical power unstated.
- Policy/practice relevance implicit — no "so what beyond this site."
- Bibliography / figures / appendices incomplete — "Figure XX" placeholders; bibliography embedded in body instead of a separate section.
- Mentoring-plan gaps that recur: a year/stretch-area with no named mentor; no contingency if a mentor leaves; mentor willingness to file annual reports not confirmed.
- Don't forget formatting compliance (font, spacing, margins, page numbers, page limits) — it's a real screen.
- Fix the most critical gap first, with surgical tracked changes confined to the affected section, then work down by impact.
Anatomy of the components
| Component | Its job |
|---|---|
| Abstract / overview | A screening document — senior staff and the full committee read it; only a couple read the full app. Land problem + timeliness + funder-fit fast. |
| Research plan | Name the practice + the specific outcome inequality; significance & focus-area alignment; aims as a logic chain; methods + analysis plan; standalone human-subjects section; feasibility; generalizability/power; policy relevance; separate bibliography + figures/appendices. |
| Mentoring / career plan | A multi-year plan to grow the scholar into new expertise — named mentor per stretch area, real meeting cadence, barriers mitigated + contingency, scholar agency, confirmed mentor commitments. |
| Budget justification | Every line tied to a research duty; community-partner & participant costs as explicit line items; real fringe rates; honest logistics. |
| Letters / biosketch | Often due earlier than the plan; mentors confirm willingness incl. reporting. |
Research-plan craft
- Frame the practice, not the problem. Open Significance by naming the practice(s) under study and the population/inequality dimension; recast the background literature as rationale, not the subject.
- Pivot to the evidentiary gap. "The literature documents the inequality but has not tested whether [practice] reduces it" — then cite each practice's evidence base and name the precise untested gap.
- Map aims to the funder's review verbs as a logic chain (e.g. build → test → increase understanding): diagnose/co-produce → co-develop/model → pilot/feasibility. Justify the type of study ("feasibility studies are indicated when no prior evaluation exists and prior work wasn't community-partnered").
- One specific, defensible innovation "first" — not "first study of X" but the first to combine these methods on this problem for this population.
- Feasibility = anticipate failure in-line. Give each aim a "Considerations & Alternative Strategies" block. The ranked-list contingency is elegant: if the top-ranked strategy is unfeasible, the ranked list supplies the next candidate without restarting the research.
- Significance = local case as national exemplar, with concrete numbers. Not
"our city faces environmental challenges" but the specific, vivid fact. State the
funder-relevant irony plainly. (This install's original example:
LOCAL.mdbeside this skill, when your copy keeps one — Significance example.) - Conceptual model belongs to research-methods (see that skill) but in the proposal it must be layered, mechanistic, with the structural cause upstream and intersectionality as a cross-cutting lens, not a box.
Mentoring / career-plan craft
The plan is judged on whether the award adds value beyond an ordinary collegial relationship. Assert (and design toward) these eight points:
- Sustained, structured multi-year mentorship with accountability.
- Protected time for risky, innovative work the award enables.
- Cross-disciplinary bridge-building (mentors outside the scholar's field).
- Access to a cohort community (horizontal peer learning).
- Career shaping, not just project advising — trajectory and research identity.
- Legitimacy & visibility (invitations, editorial boards, collaborations).
- Resources tied to the relationship (mentor co-investment).
- Explicit shared focus on reducing inequality.
Per-mentor structure that works: (1) current relationship history → (2) how the award changes it → (3) specific activities + time commitment → (4) barriers mitigated. Pair each stretch area with a named mentor (leave no year uncovered). Lead with the scholar's agency; quantify the increase the award enables ("from once a semester to recurring biweekly structured meetings"); state a real cadence with a phased transition ("monthly during planning … as needed once implementation is underway"); name barriers and how they're solved plus a contingency; use specific mechanisms (named lab, curated readings, hands-on facilitation), not platitudes; confirm mentor commitments including annual reports.
Budget-justification craft
- Tie every cost to a named research duty (e.g. "1.5 months summer salary" for specific responsibilities).
- Make community-partner and participant costs explicit line items — a named subcontract, a separate participant/honoraria line — never buried in "Other Direct Costs." Separate participant honoraria from space/materials: signals participation is compensated work (CBPR power-sharing).
- Use real institutional fringe/payroll rates; don't round or guess.
- Keep logistics honest and modest; the narrative must defend the scale of the partnership against the dollars.
Do's & don'ts
Do: audit responsiveness before prose; center the fundable unit; name the specific outcome inequality with mechanism; give every data source an analysis plan; a standalone human-subjects section; state generalizability/power and policy relevance; concrete numbers over abstractions; named mentor for every year; explicit community/participant budget lines; check the exact formatting spec.
Don't: center causes/consequences when the funder funds practices; frame the population as "vulnerable/at-risk" instead of knowledge holders; claim CBPR/YPAR without power-sharing language in every section; leave the partner vague ("we'll work with community partners"); call a list of data sources an analysis plan; leave figure/ bibliography placeholders; over-promise scale (a feasibility study's strength is rigor about what isn't yet known).
Voice
Grant prose is scholarly but active and confident. For register, word choice, and
positive reframing, pair with a scholarly-voice/academic-writing skill if you keep one
(this install's: LOCAL.md beside this skill, when your copy keeps one — Voice).
Signal-sophistication vocabulary (structural racism vs. "disparities", epistemic
justice, solastalgia, climate gentrification, allostatic load, adaptive implementation)
is used precisely, never as decoration.
Before you call this done
- Element-by-element audit against the funder's required-components list and review criteria is complete — no "Figure XX" placeholders, bibliography in its own section.
- The fundable unit (the practice/intervention), not the background problem, is the subject of Significance and the aims; the outcome inequality is named with its mechanism.
- Every data source has an analysis/interpretation framework, and a standalone Human Subjects Protection section exists (IRB, data security, confidentiality, risk, mitigation).
- Every mentoring stretch-area/year has a named mentor with a real cadence and a contingency; community-partner and participant costs are explicit budget lines.
- Formatting spec verified against the guidelines (font, spacing, margins, page numbers, page limits).
Capturing learnings
Record funder-specific rules, what reviewers rewarded/penalized, strong section
templates, and budget patterns in LEARNINGS.md beside this skill, when your copy keeps
one. Read it before starting grant work; update after each application or review round.