City design & planning

<context>
I want to improve [place, e.g. the 1.2 km main street between the
river and the rail station] in [jurisdiction, e.g. a 60,000-person
city in Oregon].
My role: [role: resident | advocate | business owner | city staff];
I'm not a planner.
</context>
 
<task>
Interview me the way a planner scopes a street or neighborhood
redesign, before any proposals. Don't propose anything yet.
</task>
 
<requirements>
- At most 10 questions, grouped: place and boundary; existing
  conditions data; rules in force (zoning, street classification,
  who owns the right-of-way); goals and metrics; people affected;
  decision, budget and timeline.
- For each question: why it matters, typical answers, and the
  official source where I'd find the data (e.g. the state DOT for
  traffic counts, the city's zoning map).
</requirements>
 
<process>
Ask all the questions in one message, then stop and wait. When I
answer, restate the brief as a list, write [NEEDS SOURCE] for every
figure I don't have and [CHECK LOCAL CODE] for every rule I haven't
pasted, and end with a list of the assumptions you made.
</process>

<materials>
[paste the brief from stage 1, the data tables, and the zoning and
street-classification text]
</materials>
 
<task>
Propose [number of scenarios: 2 | 3 | 4] materially different
scenarios for this [scale: street | corridor | neighborhood], so
[decision maker, e.g. the city council] can choose one to develop.
</task>
 
<requirements>
- The scenarios differ in ambition and approach, e.g.
  [scenario set, e.g. tactical pilot, road diet, rezoning plus
  street redesign].
- For each: street changes as a cross-section table that sums
  exactly to the right-of-way; land use and zoning changes (FAR,
  form-based standards, missing middle types); transit and walking
  changes; rough cost (low, medium or high) and timeline; who
  benefits, who pays, displacement risk; approvals needed.
- Describe LOS and VMT effects side by side, and say which metric
  the jurisdiction uses for review: [CHECK LOCAL CODE].
- Never invent numbers, names, quotes or sources: write
  [NEEDS SOURCE] where support is missing.
</requirements>
 
<output_format>
A section per scenario; then an evaluation matrix (scenarios × my
goals, scored 1-5 with one-line reasons); then a recommendation and
the assumption that would flip it.
</output_format>
 
<process>
Stop after the recommendation and wait for my choice. End with a
list of the assumptions you made.
</process>

<materials>
[paste the brief, the chosen scenario, the data tables (counts,
crashes, transit frequency, parcels), the zoning text and the
comprehensive plan policies that apply]
</materials>
 
<context>
Place: [place, e.g. Main Street from 1st to 9th, 24 m right-of-way,
a state route]. Jurisdiction: [jurisdiction, e.g. a US city of
60,000 in Oregon].
Readers: [audience: neighbors | planning commission | city council],
who [readers' knowledge, e.g. know the street well but not planning
terms].
</context>
 
<task>
Develop scenario [chosen scenario, e.g. B, a road diet with
protected bike lanes] into a recommended redesign plan with an
engagement plan, ready for [next step, e.g. a council work
session].
</task>
 
<requirements>
- Existing conditions from my data only.
- Goals as measurable objectives with indicators (baseline, target,
  year).
- Design: cross-sections per segment that sum exactly to the
  right-of-way; intersection treatments; transit stops; curb uses
  (loading, parking); trees and green infrastructure. Apply complete
  streets principles and the NACTO Urban Street Design Guide, naming
  the section relied on; mark every dimension a traffic engineer
  must confirm.
- Land use: zoning changes that would support the design (FAR,
  build-to lines, active frontage, missing middle types), each
  marked [CHECK LOCAL CODE].
- Equity: benefits, burdens and displacement risk by group, with
  anti-displacement measures.
- Implementation: phases, actors, rough cost (low, medium or high),
  approvals (state DOT, utilities, environmental review).
- Engagement plan: the IAP2 level promised at each stage, methods
  per audience, access (languages, times, childcare), how input is
  reported back; flag where it risks tokenism on Arnstein's ladder.
- Never invent numbers, names, quotes or sources: write
  [NEEDS SOURCE] where support is missing.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- Every figure is from my data or marked [NEEDS SOURCE].
- Every cross-section sums to the right-of-way; every density says
  net or gross.
- Each element says who benefits and who pays; nothing is all
  upside.
- Every action has an actor, a money source and an approval path.
- The engagement plan says what the public can actually change.
</quality_bar>
 
<output_format>
Sections: a one-paragraph summary for the readers named above;
existing conditions; goals and indicators (table); design
(cross-section tables, then intersections); land use; equity;
implementation (table: action, who, when, cost, approval);
engagement plan (table: stage, IAP2 level, methods, audiences);
data to collect before and after.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would change
your answer; otherwise go on. End with a list of the assumptions
you made.
</process>

<materials>
<brief>
[paste the brief and data tables from stage 1]
</brief>
<draft>
[paste the plan from stage 3]
</draft>
</materials>
 
<task>
Review this redesign plan against the brief, as a pre-hearing staff
review. Critique only: don't rewrite it or add proposals.
</task>
 
<requirements>
- Score each 1-5 with evidence quoted from the draft: grounded in my
  data; arithmetic (cross-sections, densities, capacities); right
  scale; rules (zoning vs. plan policy vs. street standards, quoted);
  trade-offs; equity; implementation; engagement.
- Recompute every cross-section and capacity figure.
- List every number in the draft that is neither in my data nor
  marked [NEEDS SOURCE].
- Mark rules you can't confirm from the pasted text
  [CHECK LOCAL CODE].
</requirements>
 
<quality_bar>
Before answering, check your review against:
- Every score cites the draft.
- Fixes are ranked by how much they would change the decision.
- The review adds no proposals of its own.
</quality_bar>
 
<output_format>
A rubric table; the top 5 fixes, ranked; the questions a planning
commission is likely to ask.
</output_format>
 
<process>
End with a list of the assumptions you made.
</process>

Here is the zoning text [paste district text] and the proposal [paste
amendment or staff report]. In plain language: what current rules
allow by-right on a typical lot, what the proposal changes (before and
after table) and what it would let someone build, with the arithmetic.
Quote the clause for each claim; mark gaps [CHECK LOCAL CODE].

<materials>
[paste the zoning district text and definitions, plus the proposed
amendment or the project's staff report]
</materials>
 
<context>
District: [district, e.g. C-2 downtown commercial]. Typical lot:
[lot, e.g. 15 × 40 m].
Readers: [readers: neighbors | commissioners | a reporter],
who know the area but not zoning terms.
</context>
 
<task>
Write a neutral one-page explainer of what the rules allow now and
what the [change: zoning amendment | development proposal | variance]
would change, ahead of the hearing.
</task>
 
<requirements>
- Current rules: uses, height, setbacks or build-to lines, FAR, lot
  coverage, parking, by-right vs. discretionary review.
- A before-and-after table of what changes.
- What could realistically be built on the typical lot before and
  after, with the arithmetic (floor area = FAR × lot area; homes =
  floor area ÷ a stated average home size).
- For a project: height and number of homes compared with nearby
  buildings; traffic and parking findings quoted from the documents,
  not estimated.
- How and when to comment.
- Quote the clause behind each statement and use only the pasted
  text; write [CHECK LOCAL CODE] where it doesn't settle a rule.
- Never invent numbers, names, quotes or sources: write
  [NEEDS SOURCE] where support is missing.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- Neutral: no adjective that argues for or against.
- Every rule has a quoted clause; every number is quoted or computed.
- A neighbor finds "how tall, how many homes, how much parking" in
  under a minute.
</quality_bar>
 
<output_format>
One page: a three-sentence summary; a rules table; the
before-and-after table; what could be built; how to comment. Then
an FAQ of 8 questions with neutral answers.
</output_format>
 
<process>
End with a list of the assumptions you made.
</process>

From this parcel table [paste parcel table], estimate housing capacity
under [scenarios, e.g. FAR 1.5 today vs. FAR 3.0 proposed]: floor area,
homes at a stated average size, realistic redevelopment share, and net
density in du/ha and du/acre. Show every step, label net or gross, and
mark any parcel figure not in my table [NEEDS SOURCE].

<materials>
[paste a parcel table: ID, area, current use, zoning, owner type,
constraints such as floodplain or historic status]
</materials>
 
<context>
Area: [area, e.g. within 800 m of the Main St station].
Purpose: [purpose: comparing rezonings | a housing target | a grant].
</context>
 
<task>
Estimate housing capacity under [scenarios, e.g. current zoning and
the proposed upzoning], so the scenarios compare on homes added.
</task>
 
<requirements>
- Per parcel: allowed floor area (FAR × parcel area, capped by height
  and coverage), residential share, homes at [average home size, e.g.
  70 m² plus 15% circulation], and a realistic redevelopment share
  (not every parcel redevelops: state the rate and why).
- Totals: theoretical vs. realistic capacity over [horizon, e.g. 10
  years].
- Density as net du/ha and du/acre (1 du/acre ≈ 2.47 du/ha), plus
  gross for the whole area.
- Exclude or flag constrained parcels (floodplain, historic, public).
- Never invent numbers, names, quotes or sources: write
  [NEEDS SOURCE] where support is missing and [CHECK LOCAL CODE]
  for zoning limits I didn't paste.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- Every parcel figure comes from my table.
- The arithmetic chain is shown and redone once.
- Net and gross densities are labeled and plausible for the
  building types implied.
</quality_bar>
 
<output_format>
Parcel table with computed columns; a scenario totals table; a
paragraph on the assumptions that move the result most.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would change
your answer; otherwise go on. End with a list of the assumptions
you made.
</process>

Audit this walk along [route, e.g. Elm St from the school to the
grocery] from my photos and notes [paste photos and notes]: sidewalks,
crossings, speeds, block length, shade, active frontage, seating,
destinations within 10 minutes. Score each 1-5 citing a photo, then
give the five cheapest high-impact fixes and who handles each.

<materials>
[paste or attach photos from the walk, in order, each with a note:
where, what time, what you noticed]
</materials>
 
<context>
Route: [route, e.g. Elm St from Lincoln Elementary to the grocery,
about 900 m]. Place: [jurisdiction, e.g. a suburb of Columbus,
Ohio]. Who walks it: [walkers, e.g. kids going to school, older
adults going to the bus stop].
</context>
 
<task>
Audit this route for walkability and rank the fixes, so I can take
a short, credible list to
[recipient: the transportation department | a council member].
</task>
 
<requirements>
- Criteria: sidewalk continuity and width; crossings (spacing,
  signals, sight lines, crossing length); traffic speed and lane
  width; block length; shade; active frontage; places to sit;
  destinations within a 10-minute walk.
- Score each 1-5 with evidence from a named photo.
- Reason about safety the Vision Zero way: design for human error,
  speed first.
- Fixes: cheapest high-impact first, with who in the city handles
  each; mark anything that needs a traffic engineer.
- Never invent numbers, names, quotes or sources: write
  [NEEDS SOURCE] where a photo doesn't show it, and
  [CHECK LOCAL CODE] for street standards.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- Every score cites a photo.
- The fixes serve the people who actually walk this route.
- No fix is vague ("improve crossings"): each names a place and a
  treatment.
</quality_bar>
 
<output_format>
Score table (criterion, score, evidence); five fixes (where, what,
cost low, medium or high, who handles it); data to collect next
(speeds, crashes, counts).
</output_format>
 
<process>
End with a list of the assumptions you made.
</process>

District: [land cover, e.g. 12% tree canopy, flat dark roofs, 4 ha of
surface parking]. Rank heat fixes (street trees, shade, cool roofs,
cool or permeable pavement, depaving, cooling centers) by impact,
cost, time to effect and who benefits, favoring the most exposed and
vulnerable blocks. Use only my data; list what to get next.

<materials>
[paste land cover, tree canopy %, roof types, parking area, and
income and age by block; add surface-temperature data if you have
it]
</materials>
 
<context>
District: [district, e.g. the Eastside industrial edge, about 2
km²]. Climate: [climate: hot-dry | hot-humid | temperate with heat waves].
Budget and owner: [budget, e.g. a $2 million city grant over 3
years].
</context>
 
<task>
Rank heat-island interventions for this district and say where to
start, so the grant goes to the blocks that need it most.
</task>
 
<requirements>
- Interventions: street trees, shade structures, cool roofs, cool or
  permeable pavement, depaving parking, green infrastructure,
  cooling centers.
- For each: impact, cost, time to effect (trees take years), upkeep,
  who benefits.
- Prioritize blocks by exposure (heat, canopy) and vulnerability
  (age, income, no air conditioning), using only my data.
- Never invent numbers, names, quotes or sources: write
  [NEEDS SOURCE] where support is missing.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- The ranking uses my data, not a generic list.
- Quick shade and long-term trees are both in the plan, with timing.
- Benefits are broken down by group, not mentioned in a closing
  line.
</quality_bar>
 
<output_format>
Table: intervention, impact, cost, time to effect, upkeep, who
benefits. Then the top blocks with reasons. Then the data to get
next.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would change
your answer; otherwise go on. End with a list of the assumptions
you made.
</process>

Plan a [length: 3-month | 6-month | 1-year] tactical pilot to test
[change, e.g. a plaza in the Oak St slip lane] with paint, planters
and movable seating: layout, materials and rough cost, permits and
partners, before-and-after measures, and the rule for making it
permanent. Mark permits [CHECK LOCAL CODE]; every measure has a baseline.

<materials>
[paste or attach photos and dimensions of the site, current counts
or speeds if any, and the city's pilot or encroachment rules]
</materials>
 
<context>
Site: [site, e.g. the Oak St slip lane, about 25 × 8 m of asphalt].
Sponsor: [sponsor: residents | a business association | the city].
Budget: [budget, e.g. $15,000 including volunteer time].
</context>
 
<task>
Plan a tactical urbanism pilot that tests [change, e.g. turning the
slip lane into a plaza] and produces the evidence for a decision on
a permanent project.
</task>
 
<requirements>
- Layout with dimensions; materials (paint, planters, posts,
  movable seating) with rough costs; installation day plan.
- Safety: emergency and delivery access, sight lines at crossings;
  mark what a traffic engineer must approve.
- Permits, insurance and maintenance: [CHECK LOCAL CODE].
- Measures with a baseline before installation: people counts,
  stays (Gehl-style public life counts), vehicle speeds, business
  feedback.
- A decision rule set in advance: what result makes it permanent,
  changes it, or removes it.
- Never invent numbers, names, quotes or sources: write
  [NEEDS SOURCE] where support is missing.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- Every measure has a baseline and a date.
- Access for emergency vehicles and deliveries is solved, not
  assumed.
- The decision rule is written before the pilot starts.
</quality_bar>
 
<output_format>
Layout description with dimensions; a materials and cost table; a
timeline; a measurement plan table (measure, method, before, during,
after); the decision rule.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would change
your answer; otherwise go on. End with a list of the assumptions
you made.
</process>

Write a 10-question survey on [plan, e.g. the Main Street redesign]
for [community, e.g. residents, renters and business owners, English
and Spanish]. Neutral wording, no leading or double-barreled
questions, 8th-grade reading level, under 5 minutes. Check each
question against those rules and say which decision it informs.

<materials>
[paste the plan summary or the options residents will react to]
</materials>
 
<context>
Community: [community, e.g. 40% renters, a large Spanish-speaking
population, many older residents].
Format: [format: online | paper | at events | all three].
The results feed [decision, e.g. the council's choice between three
street designs in May].
</context>
 
<task>
Write a community survey whose results can inform that decision and
be reported back honestly.
</task>
 
<requirements>
- At most [length: 8 | 10 | 15] questions, about 5 minutes.
- Each question maps to a decision or to who answered (demographics
  last, optional, with a reason given).
- Neutral wording; no leading, loaded or double-barreled questions;
  balanced answer scales with a "don't know" option.
- Plain language at about an 8th-grade reading level; translation
  notes for [languages, e.g. Spanish and Vietnamese].
- Never invent numbers, names, quotes or sources: write
  [NEEDS SOURCE] where support is missing.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- No question assumes the answer the city wants.
- Every question's results would change something in the decision.
- Someone who didn't follow the project can answer every question.
</quality_bar>
 
<output_format>
The survey, numbered; then a table: question, decision it informs,
answer type; then how to report results back to respondents.
</output_format>
 
<process>
End with a list of the assumptions you made.
</process>

Same street, same buildings, same camera position and lens as the
reference photo. Change only: [change, e.g. four travel lanes to two
plus a center turn lane], [bike lane: none | painted | protected],
trees every 8 m in the parking lane, wider sidewalk with benches.
Late afternoon, people walking, photoreal, eye level, no text.

<materials>
[paste or attach the street photo, and the cross-section table of
the proposed design]
</materials>
 
<context>
Street: [street, e.g. Main St at 4th, looking north]. The image is
for [use: a public meeting | a flyer | a grant | a council memo],
labeled as an illustration, not a design.
Generator: [generator, e.g. Adobe Firefly or a chat assistant's image
editor], from the photo.
</context>
 
<task>
Write an image-to-image prompt and an edit plan that turn this
photo into an honest illustration of the proposed cross-section.
</task>
 
<requirements>
- Keep the buildings, camera position, lens and light of the photo;
  change only what the cross-section changes, lane by lane, left to
  right.
- The widths in the image match the cross-section table; the number
  of lanes is stated explicitly.
- People, bikes and buses at realistic numbers for this street; no
  festival crowds.
- Put tool-specific parameters at the end, apart from the
  description.
- Never invent numbers, names, quotes or sources: write
  [NEEDS SOURCE] next to any tool parameter you're unsure exists.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- The prompt describes the scene, not reasons or requests.
- Nothing appears that isn't in the design (no extra plaza, no new
  buildings).
- The prompt keeps the existing signage and storefronts, so the
  place is recognizable.
</quality_bar>
 
<output_format>
The prompt; then an inpainting plan (which regions to mask, in what
order); then three things to check in the result.
</output_format>
 
<process>
End with a list of the assumptions you made.
</process>

Write a Python script with OSMnx and GeoPandas that downloads the
walking network for [place, e.g. "Portland, Oregon, USA"], computes
intersection density per km² and average street segment length, and
maps the 10-minute walkshed (80 m a minute) from [start, e.g. an
address or coordinates]. Use only documented functions; say which version.

<materials>
[paste a boundary polygon (GeoJSON) if the area isn't a named
place]
</materials>
 
<context>
Area: [area, e.g. "Eastside, Columbus, Ohio, USA" or the polygon].
Start point for walksheds: [start, e.g. Lincoln Elementary, or its
coordinates]. Skill: [skill: never used Python | some Python | comfortable].
</context>
 
<task>
Write a Python script that measures this area's walking network, so
I can compare it with other neighborhoods using real numbers.
</task>
 
<requirements>
- OSMnx and GeoPandas; say which OSMnx version you assume and use
  only functions in its documentation.
- Download the walking network; compute intersection count and
  density per km², average street segment length and a block-size
  measure; build walksheds at 5, 10 and 15 minutes (80 m a minute)
  along the network, not as circles.
- Save a GeoPackage of the results and a PNG map.
- Comment each step in one line so I can redo it in QGIS.
- Never invent numbers, names, quotes or sources: write
  [VERIFY] beside any function you're unsure exists in that
  version.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- Every function and argument exists in the stated OSMnx version.
- Distances are in meters in a projected coordinate system, not
  degrees.
- Walksheds follow streets, so barriers like highways show up.
</quality_bar>
 
<output_format>
The script in one block; then how to install and run it; then what
each output means.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would change
your answer; otherwise go on. End with a list of the assumptions
you made.
</process>

How can I make my downtown better?

<materials>
Main Street, downtown: 4 travel lanes, 24 m building face to
building face, a state route, 15,000 vehicles a day, posted 35 mph.
3 pedestrian injuries in 5 years at 2 crossings (city crash data).
Ground-floor vacancy 40% (downtown association count). Zoning C-2:
FAR 1.5, 2 parking spaces per 1,000 sq ft. One bus route, every 30
minutes. The riverfront park is cut off by a 1.5-acre city-owned
parking lot.
</materials>
 
<context>
A US city of 60,000. I'm a resident on the downtown advisory
committee, not a planner. This goes to a city council work session;
council members know downtown but not planning terms.
</context>
 
<task>
Diagnose downtown's problems and propose three options for council
to choose from: a low-cost pilot, a street redesign, and rezoning
plus a street redesign.
</task>
 
<requirements>
- Diagnosis using Jacobs's four conditions for diversity and
  Whyte's findings on public space.
- Each option: a cross-section table that sums to 24 m, following
  complete streets principles and the NACTO Urban Street Design
  Guide; housing capacity with the arithmetic; rough cost (low,
  medium or high); timeline; what needs state DOT or traffic
  engineer approval.
- Goals: fill storefronts, add 300 homes downtown in 10 years, zero
  serious injuries, connect downtown to the river.
- Never invent numbers, names, quotes or sources: write
  [NEEDS SOURCE] where support is missing; zoning and street
  standards beyond what I gave: [CHECK LOCAL CODE].
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- Every figure is from my data or marked [NEEDS SOURCE].
- Every cross-section sums to 24 m.
- Each option says who benefits and who bears the cost.
- Approvals and actors are named for each option.
</quality_bar>
 
<output_format>
Diagnosis (half a page); one section per option; an evaluation
matrix (options × the four goals, scored 1-5); data to collect next.
</output_format>
 
<process>
End with a list of the assumptions you made.
</process>

DIAGNOSIS
Jacobs: one primary use (offices, 8 to 5), so the street empties
after 6 pm; short blocks and old buildings mixed with 1970s infill
help; too few residents for evening life.
Whyte: the park has no seating toward the street and no direct
route in; the parking lot is its edge.
 
OPTION A, pilot (low cost, 6 months): paint-and-post curb
extensions at the 2 injury crossings; parklets at 3 vacant
storefronts; a weekend market on the river lot.
 
OPTION B, road diet (medium cost, 2-3 years), meters:
sidewalk 3.0 | bike 1.8 | buffer 0.6 | parking 2.1 | lane 3.0 |
turn 3.0 | lane 3.0 | parking 2.1 | buffer 0.6 | bike 1.8 |
sidewalk 3.0 = 24.0
 
OPTION C, rezoning + B (higher cost, 5-10 years): a form-based code
for C-2 (build-to line at the sidewalk, 4-5 stories, active ground
floors, no parking minimums).
Housing capacity: 12 underused parcels averaging 2,000 m² =
24,000 m². At FAR 3.0: 72,000 m² of floor area; 60% residential =
43,200 m²; at 85 m² per home ≈ 508 homes; 40% redevelop in 10
years ≈ 203. The river lot (1.5 acres ≈ 6,070 m²) as a city-led
project at FAR 2.0 ≈ 12,100 m² ≈ 142 homes. Total ≈ 345: meets the
300-home goal.
 
MATRIX (1-5)    Storefronts  Homes  Safety  River
A pilot              2         1      3       3
B road diet          3         1      4       2
C rezoning + B       4         5      4       4
 
APPROVALS: Main Street is a state route, so lane changes need state
DOT approval and a traffic study [CHECK LOCAL CODE].
ASSUMPTIONS: 85 m² average home; 60% residential share; ...

The 12 parcels averaging 2,000 m² aren't in my data. Here is the
county assessor's parcel list for downtown: [paste parcel table:
ID, area, use, land value, building value]. Redo Option C's
capacity from this table only: treat a parcel as underused where
its building value is below its land value, show the arithmetic
per parcel, and give totals with and without the river lot. If the
total falls short of 300 homes, say what FAR or redevelopment rate
would reach it, then re-score the matrix.

Follow-up critique prompts
 
- "List every number you used and where it came from: my data,
  a stated assumption, or general knowledge."
- "Argue against your recommended scenario from a traffic
  engineer's view, then a renter's worried about rent, then a
  small-business owner's."
- "Which parts need state DOT, utility or environmental review
  approval, and roughly how long do those take?"
- "What would we measure in year 1, 3 and 5 to know it worked?"
- "Rewrite the plain-language summary at an 8th-grade reading
  level without changing what it promises."