beekeeping_apiary_records v1.0.0

name: "beekeeping_apiary_records" description: "Keeping apiary records that make next season's decisions obvious: hive inspection logs that record queen status, brood pattern, stores, and temperament in the same format every visit, varroa mite counts and treatment dates tracked per hive with thresholds for action, honey super and harvest records kept by yard, queen and split lineage kept across seasons, and a winter loss review that reads the year's logs instead of guessing. Use when setting up hive records, logging an inspection, deciding whether a mite count needs treatment, tracking supers or harvest by yard, recording a split or a new queen, or reviewing winter losses." metadata: { "includeInPrompt": true }

Beekeeping Apiary Records

Purpose

Run an apiary on records, not on memory. Every hive gets the same inspection format on every visit, every mite count gets compared to a named threshold on the day it is taken, every super and every queen carries a written history, and winter losses get reviewed against the year's own logs in spring.

The failure this prevents is common and quiet. Inspections happen, notes sit in three notebooks and a phone, mite counts get remembered as "lowish", and by spring nobody can say which colonies were treated late, which queen line lost half its colonies, or whether the deadouts starved or crashed from mites. Records in one format, read back on a schedule, answer those questions while the answers can still change next season.

This skill covers record keeping and the decisions the records drive. It does not replace a treatment product label, a state apiary inspector, or a diagnosis. Suspected foulbrood or any notifiable disease goes to the state or provincial apiary inspector, and the record for that colony says so.

Workflow

1. Give every hive and every queen a stable identity

Set this up once, before the season, and never renumber mid season.

1. Yard code and hive ID: one or two letters for the yard, a number for the hive, painted or tagged on the box. Example: N-03 is hive 3 in the North yard. A hive keeps its ID when it moves yards. The move gets logged, the ID does not change.
2. Queen ID: year raised, mother colony, source. Example: Q2025-N01 means raised in 2025 from the queen in N-01. Source is one of: home raised, bought (breeder named), swarm caught, package, or unknown. Unknown is a valid entry. A guessed lineage is not.
3. Queen marking follows the international colour code by the last digit of the year she was raised: white for 1 or 6, yellow for 2 or 7, red for 3 or 8, green for 4 or 9, blue for 5 or 0. The phrase "Will You Raise Good Bees" gives the order. Colour alone never proves age, since the cycle repeats every five years. The written queen record is the authority.
4. Colony origin line: for each hive, record how it started (overwintered colony, split from a named donor on a named date, nuc bought in, swarm) and every split it has produced, with the date and the new hive ID. Frames moved between colonies get logged on both records. That is what makes later performance traceable to a donor instead of a guess.

2. Inspect in the same format every visit

Use assets/inspection-form.md, or the CSV columns written by scripts/apiary_records.py. Same fields, same order, same codes, every hive, every visit. Fill a field with "not checked" rather than leaving it blank.

Record these, in this order:

1. Date, yard, hive ID, and whether bees were flying.
2. Queen evidence, coded: eggs, larvae_no_eggs, queen_only, none, or cells. Eggs mean a queen was laying within about the last three days, so seeing the queen herself is not required on a routine visit. Utah State University Extension gives the same rule: eggs and a steady brood pattern show the queen is laying. Queen seen is a separate yes or no field, never a substitute for the evidence code.
3. Brood frames: count of frames with brood, halves count. Brood pattern, coded: solid (a compact block of capped brood with few empty cells), spotty (scattered empty cells through the capped area), scattered (no clear block, brood spread thin), or none_seen. A spotty or scattered pattern is a flag to investigate, not a diagnosis. It can come from a failing queen, disease, mite damage, or poor nutrition, and the record should describe what was seen rather than name a cause.
4. Stores frames: frames of capped honey and nectar, counted, plus pollen seen yes or no. In fall this count gets compared to the winter target set in step 5.
5. Temperament, scored 1 to 5 on this fixed scale so scores mean the same thing across visits and keepers: 1 means bees stayed on the comb, no veil reaction, no following. 2 means some bees in the air, no stings, work stayed calm. 3 means bees bumped the veil or gloves, work still manageable. 4 means stings through gloves or bees followed after closing. 5 means work had to stop or bees followed a long way. Score what happened, not what the colony is usually like.
6. Mite sample, when taken: method, mites counted, bees sampled, percent. See step 3.
7. Treatment applied: product, dose exactly as the label states, date on, planned date off, and whether honey supers were on.
8. Supers on: count, plus super IDs if supers are tracked individually.
9. Action taken and the next check: a date plus the question that check must answer, written as a question. "Recheck" alone is not a next check. "Are there eggs" after a queenless flag is.

Keep inspections short and purposeful. The record is the point of the visit as much as the look inside. Write it at the hive, not from memory that evening.

3. Count varroa on a method, and act on a threshold

Looking at bees does not measure mites. Use a repeatable sample.

1. Method: the alcohol wash (or soap wash) and the powdered sugar shake are the two most accurate field methods, per the Honey Bee Health Coalition guide Tools for Varroa Management. A standard sample is about 300 bees, roughly a half cup, taken from a brood frame, after checking the queen is not on that frame. The alcohol wash kills the sample and gives the firmest count. The sugar shake returns the bees alive and reads a little less precisely. Pick one method as the yard standard and record which one each count used, because counts from different methods are not perfectly comparable.
2. Math: percent equals mites counted divided by bees sampled, times 100. Nine mites in 300 bees is 3.0 percent, also stated as 3 mites per 100 bees. Run it with the script: python3 scripts/apiary_records.py mite --mites 9 --bees 300 --season late_summer
3. Schedule: sample every colony at least in spring, mid summer, and late summer. USDA ARS guidance is to check monthly through summer and about every three weeks in fall, weather permitting. A colony with no fall sample is an unknown going into winter, and the winter review will say exactly that.
4. Thresholds, the action rules for this skill. USDA ARS recommends treatment when a wash or shake reaches 3 mites per 100 bees, or when a sticky board drops more than 30 mites per day:

• Below 1 percent: monitor. Retest on the normal schedule.
• 1 to 2 percent: monitor in spring and early summer, retest in about four weeks. In late summer and fall, retest in two to three weeks and have a treatment chosen.
• 2 to 3 percent: in late summer and fall, treat now. The bees raised in that window are the winter bees, and mite numbers climb fast there. Earlier in the season, treat, or retest within two weeks and treat if the count is still at that level or rising.
• At or over 3 percent: treat now, in any season, and book the post treatment sample at the same time. A rising trend across two samples counts even under a threshold. A colony that went 0.3, then 1.3, then 2.3 percent is on a line that crosses the threshold during the worst part of the season.

1. Treatment record and choice: choose the product with the five questions in the Honey Bee Health Coalition decision tool: is the count over threshold, synthetic or naturally derived preference, is brood present, are honey supers on, and is the colony population rising, at peak, falling, or dormant. Record the answers with the treatment row. The product label is the law for dose, temperature limits, super rules, and withdrawal. USDA ARS lists the EPA approved options it names as amitraz, formic acid, HopGuard, oxalic acid, and thymol. Rotate product class between treatments so the same chemistry is not used back to back, which slows resistance.
2. Prove it worked: sample again after the treatment window closes, about two to four weeks after application ends for most products, following the label where it states a check. A treatment with no follow up sample is recorded as unverified, and the winter review treats it that way.

4. Log supers and harvest by yard

Honey records live at yard level, with hive level notes where they matter.

1. Super log per yard: super ID, hive it went on, date on, date off, frames capped at removal. A super that moves between hives gets both moves logged.
2. Harvest record per yard batch: date extracted, supers in the batch, total extracted weight, and the yards it came from if a batch mixes yards. Jars or buckets sold or given away can be tallied against that batch total, so the end of season number is a subtraction, not a memory.
3. Never treat with a product whose label forbids use with supers on, and never guess whether supers were on. The inspection row already carries supers_on for exactly this check.

5. Close the season: fall status and the winter target

At the last fall inspection, every colony gets a closing row with: mite percent and sample date, treatment status (treated and verified, treated and unverified, or not treated, with the reason), stores frames, queen age and ID, temperament for the season, and status going into winter.

Set a winter stores target for your climate before that inspection and write it on the form. Guidance varies by climate, which is the point: University of Florida IFAS guidance for northern climates puts the need at about 90 pounds of honey reserves, while guidance for milder areas commonly puts it at 60 to 80 pounds. A deep frame of capped honey holds roughly 6 to 8 pounds as a working estimate, so convert the target to frames for your equipment and record the frame target, not just the weight. Colonies under target get fed in fall while syrup is still taken, combined if they are also weak, or flagged for emergency winter feed, and the choice is written on the closing row. Remove queen excluders before winter, note mouse guards and insulation or wind protection fitted, and record all of it on the same row.

For context when reading your own loss numbers later: the US Beekeeping Survey run by Auburn University, Oregon State University, and the Apiary Inspectors of America estimated a 30.3 percent winter loss for winter 2025 to 2026, and beekeepers in that survey said about 21 percent would be acceptable to them. Your yard number is yours to judge, but it is not judged in a vacuum.

6. Review winter losses from the logs

Run assets/winter-review-form.md once in early spring, after the first full inspection of every colony that went into winter.

1. Count first: colonies into winter, alive, dead, alive but queenless or too weak. Loss percent is dead divided by into winter, times 100. Use the script review mode on the season log to get visits, last mite result, and dead flags per hive: python3 scripts/apiary_records.py review --file yard-log.csv
2. For each dead colony, read its own rows before opening the deadout: last fall mite percent, treatment and its dates, fall stores, queen age and source. Then record what the deadout shows. Bees head first in empty cells with no stores in reach points to starvation. Stores still present with a high last mite count and late treatment points the other way. If the rows needed to decide were never written, the cause is recorded as unknown from records. That entry is useful. It names the record to fix next season.
3. Compare groups, not single hives: mean fall mite percent in colonies that lived against colonies that died, treated on trigger against treated late, at stores target against under target, loss rate by queen age group, and loss rate by queen source. Small yards give weak statistics, so read a one season pattern as a prompt to watch, and a two or three season pattern as a basis to change practice.
4. Write decisions as dated actions: fixed sample dates for next season, a treatment timing change, a stores target or feeding change, queen lines to raise from and lines to stop raising from, and any yard or equipment change. A review that ends in general intentions did not happen.

A full season worked through this system, spring inspection to winter review, is in references/worked-season.md.

Output Contract

A season run with this skill produces:

• One inspection log per yard (CSV from scripts/apiary_records.py, or the paper form in assets/inspection-form.md transcribed), with every visit in the coded format from step 2 and no blank fields.
• A varroa tracker inside that log: per hive, per sample, the method, count, sample size, percent, the threshold action taken, treatment product and dates, and the post treatment sample result.
• A super and harvest log per yard: supers on and off by date and hive, batch extraction weights, and batch totals.
• A lineage ledger: one line per queen (ID, year, mother, source, colonies she has headed) and one origin line per colony (how it started, donor where split, splits produced).
• A fall closing row per colony with mite status, treatment status, stores against the written target, and winter preparations.
• A completed winter review per yard in spring: counts, loss percent, per dead colony findings, group comparisons, and dated decisions for next season.

Operating Rules

• Same format every visit. A new field is fine, a field that changes meaning mid season is not. If the format must change, change it between seasons and note the date.
• Record evidence, not conclusions. "No eggs, brood scattered, cells present" is evidence. "Queen failing" is a conclusion that the next keeper, or the same keeper in spring, cannot check.
• The label is the law. This skill never states a dose, a temperature range, or a super rule for a product. The treatment row records the product and that the label was followed.
• A mite count without a method and a sample size is not a count. Do not compare it to a threshold and do not average it with real samples.
• Never invent a backdated record. A missed inspection is logged as missed when it is discovered. Reconstructed numbers poison every comparison the winter review makes.
• Suspected foulbrood, a notifiable disease, or a colony pattern you cannot explain goes to the state or provincial apiary inspector. Record the referral and the date on that colony record.
• Local rules govern treatments, hive registration, and movement of bees and used equipment. Check the rules for the state, province, or country the yard sits in. Thresholds and product approvals in this skill reflect US sources.
• Keep the logs where they survive a lost phone: one paper or file copy at the yard or home base, one copy backed up elsewhere, reviewed at the winter review.