Ok, the problem with writing proposals is that only about 20% of National Science Foundation proposals are funded, which means that 80% are rejected. Some part of the rejections are by people who don't know how to write a proposal and who will never be funded in 10,000 years. So, the odds may be slightly better for those who can write proposals (I've had 1 NSF proposal funded as the only person on it, and about 5 USGS-NEHRP proposals funded; otherwise my salary is from bigger projects where I have not been lead (plus some small budget SCEC proposals, but not for the last 4 years).

Scientific writing might be a little like a historical novel or a non-fiction work: you have to read a lot of information. You also have to do your own research. I always make about 4 or 5 pages of high-quality graphics with captions out of the 15 page proposal. Too bad I don't take the time to work on figures on my very out-of-date web site, and link to them here.
In addition to the 15 pages, you get a 1 page summary or abstract. You have to grab the attention (why it is important: is called "Intellectual Merit"), say what you propose to do, and how you will do it. You are Required to have a section in the Summary and in the main proposal called "Broader Impacts". You include things like student training and support (can be salary for students), "outreach" (communication to the public), and something that is important that might not be part of the Intellectual Merit. A common one for us would be earthquake hazard. For example, we have apending proposal to studeny the offshore fault system of Haiti, where the Intellectual Merit is understanding oblique contractions, including the offshore fold and thrust belt.
Broader Impacts would be intellectual infrastructure(?): we would bring a Haitian student to the heartland to train him/her in these kinds of studies. It is badly needed to have Haitian expertise, by Haitains living in Haiti.

There is all kind of additional information that goes with a proposal: Budgets, Budget explanation, Facilities, 2 page CVs in a particular format, Current and Pending support for the P.I.s, (no, not private eyes: Principle (pal?) Investigators). Results from prior NSF support is required for each P.I. within the 15 page limit, unless you never have had such support as a P.I.

OK, this post is long enough; I'll post more here later on what goes into a manuscript for publication.

Chris