site-analysis

Reads a site for the things that will decide the building rather than for a set of diagrams: what is fixed and cannot move, what the approach and the ground do to where a building can sit, where the s

SKILL.md

GitHub

Site Analysis

Site analysis usually produces a sheet of diagrams that nobody refers to again: a sun path, a wind rose, a noise arrow. The useful version answers a different question. Not what is on the site, but what the site has already decided, and what still has to be found out before anyone commits.

Scope and limits — read first

What this does. Works the site for its decisions: what is immovable, where a building can physically go, what the approach and the ground allow, what the context requires, and what is unknown. Keeps observation and inference separate, and ends with a list of questions rather than a set of diagrams.

What this does not do.

  • It collects no data. climate-zone-mapper, seismic-lookup and wind-snow-lookup record site values; zoning-summary and /test-far hold the zoning; lookup-flood-zone the flood designation; survey-review and geotech-report-review read the surveyor's and the engineer's documents. This reads what they found for what it means to the design.
  • It is not a survey. Nothing here is measured.
  • It does not test feasibility against zoning arithmetic. test-fit does.
  • It does not design. parti-study takes what this produces.
  • It does not replace the surveyor, the civil engineer, or the Architect of Record.

What you need before starting. The site: an address, a survey if there is one, and a visit if one has happened. The programme, or an idea of it. Whatever data has been collected. Whether the site is bought, optioned, or being considered.

Workflow

Step 1. Establish what is fixed, before looking at anything else

Ask, in one message:

  1. What is fixed and cannot be moved? Boundaries, easements, rights of way, existing structures to remain, protected trees, a required frontage, an existing access point.
  2. What is fixed by others but might be negotiable? A curb cut, a utility connection, a party wall agreement, a setback with a variance path.
  3. What data exists already, and from what document?
  4. Has anyone visited, and what did they notice that no drawing shows?
  5. Is the site committed, or is this part of deciding whether to buy?
  6. What is the programme, roughly, and how big is it against the site?

Question 5 changes the exercise. For a site being considered, the job is to find the thing that kills it, fast. For a site already bought, the job is to find what it decides.

Step 2. Work the ground and the approach

These two decide where a building can go more than anything on a sun diagram:

  • Levels. Where is it high and low, and by how much? What does that do to the entrance, the accessible route, the parking, and where a basement is possible or unavoidable? Point at civil-drawing-review.
  • What is under it. Rock, water, fill, contamination, a previous building. Point at geotech-report-review, and say plainly if no report exists.
  • Drainage. Where does water arrive from, cross, and leave? What is downstream and who owns it?
  • Approach. How do people and vehicles reach it, from which direction, and what do they see as they come? Point at site-access-study.
  • Services. Where does each utility come from, at what capacity, and what does connecting cost? This is a common late surprise on rural and infill sites both.

Step 3. Work the sun, the weather and the outlook, as decisions

Not as diagrams. For each, say what it decides:

  • Sun. Which faces get it, when, and which parts of the site are in shadow from what. Then: which rooms want it and which do not, and whether the site's good orientation and its good view are the same direction. They usually are not, and that conflict is a design decision worth naming now. Point at shadow-study and daylight-study.
  • Wind and weather. Prevailing direction, driving rain, snow drift and where it lands, and what that does to entrances and outdoor space.
  • Outlook. What is worth seeing, what is not, and what will change. A view over a neighbouring site is a view until that site is built on. Point at view-analysis.
  • Noise and nuisance. Roads, rail, flight paths, plant, a neighbour's use, and what time of day each matters.

Step 4. Work the context and the neighbours

  • What the surrounding buildings do: their height, their setback, their materials, where their windows and entrances face.
  • What overlooks the site and what the site would overlook. Privacy runs both ways and it is the most common objection.
  • What the neighbours will care about: light, view, traffic, noise, construction. Point at entitlement-tracker if approvals are discretionary.
  • Whether there is a pattern the site sits in, and whether following it or breaking it is a choice anyone has made deliberately.
  • Anything protected: heritage, trees, habitat, a viewshed.

Step 5. Separate what was observed from what was inferred

Every finding above gets marked:

Finding Observed / From a document / Inferred Source

Inference is where site analysis goes wrong. A wind direction assumed from the region, a soil condition assumed from the neighbourhood, a service capacity assumed from a nearby building: each is reasonable and each has broken a scheme. Count them.

Step 6. Report

  • What kills the scheme, if anything. First, and plainly. On a site being considered this is the whole report.
  • What is fixed, and what is negotiable with whom.
  • What the ground and the approach decide.
  • The sun-and-view conflict, if there is one.
  • What the neighbours will raise.
  • The observed / document / inferred counts.
  • What has to be found out, ordered by what it unblocks, with who answers each and how long it takes. A survey, a geotechnical investigation, a utility enquiry, a pre-application meeting.

Close by saying nothing was measured or tested here.

Step 7. Save

Ask where. It is read back at every stage, and again when somebody asks in a year why the building sits where it does.

Rules

  • Ask whether the site is committed before anything else. It changes the exercise.
  • Work the ground and the approach before the sun.
  • Every finding is marked observed, from a document, or inferred.
  • Name the conflict between the good orientation and the good view, where there is one.
  • End with what has to be found out, ordered by what it unblocks.
  • Never collect data here. Name the skill that does.

Anti-patterns

  • Producing a sheet of diagrams that decides nothing.
  • A sun path with no statement of what it means for the plan.
  • Assuming soil, wind or service capacity from the neighbourhood.
  • Reading the site and not the approach to it.
  • Ignoring what the site overlooks, which is half the privacy question.
  • Treating a view over an empty neighbouring site as permanent.
  • Leaving the unknowns unlisted, so they are discovered in sequence rather than commissioned at once.

Resources

None. This skill is one file. Output is written directly at the path you choose.