grant-section-drafterlisted

# Grant Section Drafter ## When to Use Trigger this skill whenever the request involves drafting extended prose that carries institutional weight — grant sections, manuscript sections, cover letters, NIH-style Specific Aims, addenda, progress reports, or response-to-reviewer text. The test: would a program officer, journal editor, or search-committee member eventually read this? If yes, use this skill. If the output is a chat reply or an internal summary, do not use it. The skill is also activated automatically by the `nightly_grant_drafter` job (`agent/jobs/nightly_grant_drafter.py`), which reads the researcher's linked artifacts overnight, identifies the next unfinished section via a gap-finder call, and invokes this skill's principles to draft that section. --- ## Voice Principles All prose produced under this skill must match the researcher's published writing register. The governing rules are drawn directly from `_DRAFTER_SYSTEM` in `agent/jobs/nightly_grant_drafter.py` and the `SCIENTIST_MODE` constant in `agent/jobs/__init__.py`: **Direct, quantitative, concrete.** Claims need numbers, not adjectives. Instead of "we have extensive experience with karyotype analysis," write "we have assembled and curated eight open karyotype databases totaling 85,000+ records across six major taxonomic groups." Vague abstractions are rejected at draft time, not during review. **No corporate register.** The prose must not sound like a consulting deck or a software company's marke