Airtightness Tracker

An airtightness target is set once, early, by someone who will not be on site, and tested once, late, when fixing it costs the most. Everything between those two moments is where the number is won or lost, and almost none of it is written down anywhere.

Workflow

Step 1. Record the target and where it came from

Ask, in one message:

1. What is the target, in what units, tested to which standard, and where is it written? A number without all four is logged as incomplete.
2. Is it a contractual requirement, a certification threshold, or an aspiration? These behave differently when the project gets tight.
3. Is there an intermediate target as well as a final one?
4. What happens if it is not met? Is there a remedy in the contract?
5. Has the enclosure boundary been drawn as a single continuous surface?
6. What phase is the project in?

Question 2 is the one that predicts the outcome. An aspiration and a contractual requirement look identical in the specification and behave nothing alike in month fourteen.

Step 2. Record the boundary as a surface, not a concept

Log which face of which assembly the air barrier is on, for every enclosure type, and where the boundary changes from one plane to another. Record any volume whose inclusion is undecided: unconditioned basements, attics, adjoining tenancies, shafts, the space behind a service riser.

An undecided volume is logged as open. It changes both the target's denominator and who has to seal what, and it is usually decided by accident on site.

Step 3. Log the junctions with an owner each

#	Junction	Boundary crosses	Owner	Detail ref	Status	Date	Note

Status is one of: not detailed, detailed, submitted, installed, inspected, sealed and signed off. Never invent a status; where nobody has said, the status is what it was and the note records that it was asked.

Every junction has one owner, named as a trade or a party. A junction with two owners has none, and it gets flagged as such rather than logged as covered.

Step 4. Log every test, every inspection, and every change

Three kinds of entry, all dated:

• Test. Who tested, when, to what standard, what the result was in the target's units, and what the report says about the conditions. Record the report reference. If the result is in different units from the target, record both as given and flag the conversion as not performed here.
• Inspection. Who looked, at what, when, and what was found open. Smoke testing and thermal imaging get logged here, not as tests.
• Change. Anything that moves the target, the boundary, a junction detail, or an owner. Record what changed, who decided, and the date. A change with no decision-maker recorded is logged as such.

Step 5. Read back what is open

On request, and at every phase change, report:

• Junctions by status, with not detailed and no owner first.
• Volumes still undecided.
• The gap between the last test result and the target, in the target's units, stated as a measurement and never as a prediction of the final result.
• Changes since the last read-back.
• Whether the target itself has moved, and who moved it.

Step 6. Write the log

One dated file per project, appended to rather than rewritten. Ask where it lives on the first run and use the same path after. Every entry carries a date and a source. There is no template and no script.

Rules

• Never supply a target, a unit conversion, or a threshold. All come from the specification or the certification, quoted with the standard.
• A number without units and a test standard is logged as incomplete, never as a target.
• Every junction has exactly one owner, or is flagged as having none.
• A predicted result is never logged as a result.
• Record what someone said, with who and when. Never infer a status.
• Report not detailed and no owner first in every read-back.
• Never convert between units. Record both figures as given and flag it.

Anti-patterns

• Recording a target without its test standard, which makes two projects' numbers look comparable when they are not.
• Converting a result into the target's units so the comparison looks clean.
• Logging a junction as covered because two trades were both mentioned.
• Treating a thermal image or a smoke test as a test result.
• Predicting the final result from an intermediate one.
• Letting the enclosure boundary stay a concept rather than a named surface.
• Rewriting an earlier entry when the target changes. Log the change; keep the history.
• Running air-barrier-trace from inside this skill. It is a separate audit and its output is this log's input.

Resources

None. This skill is one file. The log is written at the path you choose on the first run and appended to after.