Condensation Check

Give it the layers in order and the conditions on both sides, and it computes the temperature at every boundary, then tells you which boundary is cold enough for the indoor air to give up its water. The heat-flow and psychrometric relations are the only things it knows. Every material property is yours.

Workflow

Step 1. Take the assembly in order

Ask, in one message:

1. The layers, inside face to outside face, in order, named. Order is the whole thing: the same layers in a different sequence give a different answer.
2. The R-value of each layer and where it came from. Air films are layers too; if you have no value for them, say so and they will be carried as unverified rather than assumed.
3. The permeance of each layer, with the unit (US perms, ng/Pa·s·m²) and the test method (dry cup, wet cup) if you have it. A layer with no figure is carried as unknown, never as vapor-open.
4. Outdoor design temperature, indoor temperature, indoor relative humidity, and the source of each.
5. Which season. Winter, heat flowing out, is the usual run; a cooling-dominated climate with a vapor-tight interior finish needs the summer run as well.

Read the layer list back, numbered, before computing anything. An assembly entered outside-in produces a clean-looking table that is exactly backwards.

Step 2. Compute the boundary temperatures

Total R first, term by term, then each boundary, with the formula printed on the line above the arithmetic:

R_total       <sum, shown term by term>
boundary n    T = T_out + (R accumulated outside of n / R_total) x (T_in - T_out)

Print the accumulated R beside each boundary temperature so any line can be re-derived. State in the output that this heat-flow relation is one of the two things the skill supplies itself.

Step 3. Compute the dew point of the indoor air

From the indoor temperature and relative humidity, using the published psychrometric relation. Print the formula, then the arithmetic, then the approximation it carries. If you were given a dew point directly, use it and say the derivation was skipped.

Step 4. Compare, and name the first condensing plane

Boundary	Temperature	Dew point	Margin	Condenses

Walk from the warm side out and name the first boundary at or below the dew point. That is the condensing plane, and it is where the water shows up. If no boundary falls below, say so and name the smallest margin instead.

Report the margin as a number and stop there. The skill holds no safety threshold. How much margin is enough is your judgement or your code's, and saying a two-degree margin passes would be supplying a value nobody gave.

Step 5. Walk the permeance in order

Read the permeance column from inside to outside and report three things:

• Where the lowest-perm layer sits relative to the insulation. A vapor-tight layer on the cold side stops outward drying and puts the condensing plane against itself.
• Whether two low-perm layers have the insulation between them. Then nothing that gets in can leave in either direction.
• Which direction the assembly can dry. State it as a direction, not a rate: there is no drying-rate model here.

Flag these, do not rule on them. Which retarder class each layer falls into, and which class your climate zone permits, is code text you supply.

Step 6. Report

The thermal table, then the vapor table, then prose: where the condensing plane lands, what moving the insulation or the retarder would do to it, and what the permeance order says about drying. Count the unverified inputs on the face of the output.

Close every run with the standing limits: steady state, one dimension, framing not in the arithmetic, air leakage not modeled.

Step 7. Save, if asked

Ask whether to write the check to a file and where. Write it directly, formula lines included. There is no template and no script.

Rules

• Never supply a material property. Not an R-value, not a permeance, not in an example, not as a placeholder that looks like a number.
• Print the formula before the arithmetic, every time. A boundary temperature nobody can re-derive is not checkable.
• Read the layer order back before computing.
• Never convert units silently. If R-values arrive in one system and temperatures in another, stop and ask which system to run in.
• A margin is a measurement. Whether it is enough is a judgement you or your code makes, and it is never asserted here.
• Say on every run that air leakage is not modeled.
• Every value carries its citation or the words not cited. There is no third state.

Anti-patterns

• Recalling a per-inch R-value for a common insulation because it is "standard". Manufacturers differ, products change, and the figure typed in is the one that has to be defensible.
• Recalling the ratio of exterior to cavity insulation a code requires, or the retarder class a climate zone permits. Both are code text and both go stale.
• Quoting a clause found by web search. Search identifies which document governs; it does not verify what the document says.
• Reporting the assembly as clear when the permeance column was never supplied.
• Presenting the result as covering the whole wall when the framing was never in the arithmetic.
• Running the numbers on an assembly listed outside-in.

Resources

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