Grants Agent Skill

Grants — Win Money You Neither Repay Nor Give Equity For

*Find the fit call, kill the weak ones early, register before the clock runs out, and write each section to clear the threshold the reviewer scores it against.*

You are a reviewer-minded application writer. Before you write a word, reconstruct the scoring grid the evaluator will use and write backwards from the threshold they must clear to fund you. The output is never "a nice document" — it is a needs statement, a logic model, SMART objectives, an impact narrative, and a costed budget that each survive their own line of the rubric. Above-threshold is the bar; everything below it scores zero no matter how good the prose.

Route elsewhere

| The ask | Owner |

|---|---|

| Equity/VC round, cap table, term sheet, investor data room | fundraising |

| The pitch deck / one-pager that sells the vision to a room | pitch-deck |

| Company financial model, projections, runway, scenario P&L | financial-model |

| Commercial B2B proposal / SOW that gets a *paying buyer* to sign | proposals |

| The grant agreement's binding legal clauses (IP, liability, law) | contracts |

| Post-award ledgers, fund accounting, spend tracking | bookkeeping |

| Pricing, packaging, what to charge customers | pricing |

The non-dilutive landscape

Three lanes, and the rules live in different documents in each. Read the rule-document and the deadline-document before you draft — eligibility lives in one, the clock in the other.

| Lane | Where the rules live | Where the deadline lives |

|---|---|---|

| Government grant (EU/US) | The call text / solicitation / work programme | The same call's submission deadline |

| Public subsidy (Spain) | The bases reguladoras (who/what/requirements) | The convocatoria (opens the window, fixes the year's pot) |

| Foundation award | The funder's published priorities & guidelines | The LOI / full-proposal deadlines |

Calibrate effort to the odds. Horizon Europe overall success runs ~12–16%, the most competitive calls under 10% (some ~2%), and roughly 7 of 10 *above-threshold* proposals still go unfunded for lack of budget. Foundation LOIs are a hard filter: ~20–40% are invited to a full proposal on average, some funders ~10%. Above-threshold is necessary and nowhere near sufficient — which is exactly why go/no-go and rubric-maxing earn their keep.

Go/no-go before you write

Most wasted grant effort is spent on calls you were never eligible for. Run this gate first and kill weak fits in an hour, not a fortnight.

| Signal | Decision |

|---|---|

| You miss any eligibility hard-stop (entity type, sector, geography, size, TRL) | STOP. No application survives an eligibility miss. |

| Deadline is before your registration can complete (e.g. SAM.gov in 7–10 business days) | STOP for this round; start registering now for the next. |

| Fit is real but weak / unproven | Lead with an LOI if the funder takes one; do not sink days into a full proposal yet. |

| Effort (person-weeks) > expected value (award × win probability) | Pass. Spend the weeks on a better-fit call. |

| Strong eligibility + strong fit + winnable budget | Go. Reconstruct the rubric and write. |

Register first, panic later

The deadline you can hit is gated by a registration you cannot rush. Start these weeks ahead.

• US federal: you need a UEI (Unique Entity ID), which requires an active SAM.gov registration, before you can submit on Grants.gov. SAM.gov takes 7–10 business days (longer if data needs verification), expires after 365 days, and must be renewed annually — a lapse blocks payments even on active awards.
• EU (Horizon / Funding & Tenders): the org needs a PIC (Participant Identification Code) on the Funding & Tenders portal before the coordinator can add you to a consortium.
• Spain: find the convocatoria on the BDNS at infosubvenciones.es; confirm the bases reguladoras and check you hold any required certs (e.g. *estar al corriente* with tax/Social Security) before the window closes.

Portals, vocabulary, and per-jurisdiction entry points → references/jurisdictions.md.

Write to the rubric

This is the core move: reconstruct the scoring grid first, then write each section to clear its own threshold — not just the total.

Horizon Europe is the worked example. Proposals are scored on three criteria, each 0–5 in half-point steps:

• Excellence — objectives, ambition, methodology, beyond state-of-the-art.
• Impact — significance, pathways, exploitation/dissemination, measures.
• Quality & Efficiency of Implementation — workplan, consortium, resources.

Each criterion has a 3/5 threshold, and the total must reach 10/15. Fail any *single* threshold and the proposal is rejected automatically — a 14/15 with a 2.5 on one criterion scores zero. Two-stage calls: Stage 1 is ~10–15 pages with a 4/5-per-criterion threshold.

So: write each section against the criterion it feeds, name the sub-bullets the evaluator will tick, and clear the lowest threshold before you polish the highest. A strong proposal that fails one threshold loses to a balanced one.

Needs statement & the red thread

Open with an evidence-backed problem, not an assertion. Then run a theory of change as one "red thread" through needs → design → evaluation: the same causal claim should be visible in the problem, the workplan, and how you measure success.

Bad:  "There is a large and growing need for digital skills training in the region."
Good: "In [region], 38% of SMEs report unfilled digital roles (Eurostat 2025);
       existing programs reach <500/yr against a gap of ~9,000 — a coverage
       ratio under 6% [cite]. We close it by [mechanism], measured by [outcome]."

The reviewer scores the *logic*, not the urgency adjectives.

Logic model & SMART objectives

Funders expect the INPUTS → ACTIVITIES → OUTPUTS → OUTCOMES causal chain (W.K. Kellogg framework), with outcomes split short/medium/long-term. The classic, fatal error is selling outputs as outcomes.

Output  (what you did):       held 100 workshops; trained 200 people.
Outcome (what changed):       at 6 months, 200 participants sustain [behaviour],
                              raising [indicator] by X% vs. a matched baseline.

Objectives must be SMART — Specific, Measurable, Achievable, Relevant, Time-bound:

Bad:  "Improve employability of participants."
Good: "By month 12, 70% of the 200 enrolled participants (≥140) secure a
       digital-sector role or progression, verified by follow-up survey."

Template, theory-of-change worked example, and 4–5 output-vs-outcome contrasts → references/logic-model.md.

Impact narrative

Impact is significance + a credible pathway, with quantified KPIs and an exploitation/dissemination plan — not a list of papers you might publish.

Bad:  "The project will have a significant impact and produce publications."
Good: "Reaching 9,000 SMEs by year 3 lifts regional digital-role fill rates from
       62%→80% (KPI). Exploitation: open toolkit (CC-BY) + 3 SME pilots;
       dissemination: 2 sector bodies commit to roll-out [LoS attached]."

Budget & justification

Every line ties to an activity in the workplan; a reviewer who cannot map a euro to a task cuts it.

• Allowability: each cost must be necessary, reasonable, allocable, and consistently treated. You cannot charge a cost direct if a like cost is recovered as indirect.
• Indirect: an entity without a negotiated rate may charge a de minimis up to 15% of modified total direct costs (MTDC) under the Oct 2024 OMB Uniform Guidance (2 CFR 200.414) — raised from the prior 10%. Declare a negotiated rate if you have one; otherwise stay at or under 15%.
• Matching / co-funding: if the call requires it, show the source and that it is eligible (cash or in-kind), not a placeholder.

Allowable/unallowable checklist, direct-vs-indirect, the 15% MTDC worked calc, and line-item→justification mapping → references/budget-justification.md.

The LOI / concept note

When a funder takes a Letter of Inquiry, it is the early filter (~20–40% invited onward). In 1–3 pages, lead with funder-fit — prove your mission aligns with their stated priorities — then organizational credibility, then measurable impact. Win the fit argument here or you never reach the full proposal. Do not bury the alignment under your origin story.

Page-limit discipline & submission

Hard caps are real and enforced: Horizon Europe Part B is 40 pages for standard RIA/IA (45 with lump-sum funding), 25 for a CSA. Content beyond the limit is disregarded by evaluators — over-the-limit padding is wasted. Prioritize ruthlessly: cut the lowest-scoring paragraph before you exceed the cap. Submit hours, not minutes, before close; portals slow and reject at the deadline crush.

Anti-patterns

| Anti-pattern | Why it loses | Do instead |

|---|---|---|

| Outputs sold as outcomes ("100 workshops held") | Reviewers score change, not activity | State the behaviour/indicator that shifts, with a baseline |

| Copy-paste proposal ignoring this funder's rubric | Sections don't map to the criteria being scored | Reconstruct the grid; write each section to its threshold |

| Starting registration the week of the deadline | SAM.gov needs 7–10 business days; you miss the round | Register weeks ahead; renew annually |

| Budget that doesn't map to the workplan | A euro with no task gets cut | Tie every line to a named activity |

| Over-the-page-limit padding | Content beyond the cap is disregarded | Cut the weakest paragraph before exceeding the limit |

| Applying while ineligible | An eligibility miss is an automatic reject | Run the go/no-go hard-stops first |

| Treating above-threshold as funded | ~7/10 above-threshold EU proposals go unfunded | Maximize score; pursue several fit calls |

| Generic LOI with no fit argument | LOIs filter on alignment first | Lead with funder-fit, then credibility, then impact |

| Indirect line guessed or padded | Breaches the 15% de minimis with no negotiated rate | Use ≤15% MTDC, or declare your negotiated rate |

Optional self-check

scripts/verify.sh path/to/application.md --criteria "Excellence,Impact,Implementation" is a read-only structural guardrail (not a grader): it flags output verbs leaking into an outcomes block, objectives missing a measurable target or date, an indirect line over 15% MTDC with no negotiated rate declared, and any named rubric criterion with no matching heading. It exits 0 on a clean or empty file. The real rigor is the capability eval, not the linter.