Period Tracker

All data lives locally in ~/Clawic/data/period/ (log-template.md for file formats, setup.md on first use). If you have data at an old location (~/period/ or ~/clawic/period/), move it to ~/Clawic/data/period/. Never sync it, never mention cycle content outside a session she opened. This skill acts directly (logs, predicts, flags); it advises, it does not diagnose.

Configuration

User-dependent variables. Defaults apply until the user states a preference; store them in ~/Clawic/data/period/config.yaml.

Preference areas to record as she reveals them:

• wording — plain vs clinical vocabulary and how she refers to her own body; affects every reply, never the stored data
• flow scale — how she reports flow (light/medium/heavy, product counts, or mL); affects logging prompts and how her reports map to the heavy-flow thresholds
• proactivity — surface pattern changes unprompted vs only on request; Red Flags are always surfaced regardless
• off-limits topics — subjects never to raise unprompted (fertility, weight, pregnancy); recorded once, honored everywhere

When To Use

• She reports a period start, flow, or symptom and wants it recorded
• She asks when her next period, fertile window, or ovulation is likely
• She wants a cycle classified (regular, irregular, heavy, long) against real thresholds, not "28 days"
• Her period is late or missed and she wants to reason through it
• Bleeding changed after starting or stopping contraception, or cycles are shifting (teens, postpartum, perimenopause)
• Not for: prenatal tracking after a confirmed pregnancy, diagnosing PCOS/endometriosis, or choosing a contraception method — route to pregnancy or a clinician

Quick Reference

Depth on demand: symptoms.md trackable catalog and phase clustering · fertility.md conception and avoidance signals · irregular.md late, missed, and shifting cycles · contraception.md bleeding on each method · pain.md cramp relief and escalation · pms-pmdd.md cyclical mood tracking · privacy.md storage, export, deletion · log-template.md data formats · setup.md first use.

Core Rules

1. Day 1 is the first day of full flow, not spotting. Prediction math anchors on Day 1; counting spotting as Day 1 shifts every downstream date by 1-2 days.
2. Predict from her median, never a textbook constant. next start = last Day 1 + median(last prediction_window cycle lengths). Median resists the one outlier cycle a mean would drag. Below 3 cycles there is no baseline: say so.
3. Attach a range, not a point. range = ± (longest − shortest)/2 over the window, rounded up. A woman with cycles 26-34 gets "predicted day X, plus or minus 4 days," never a false-precision single date.
4. Ovulation is back-counted, not forward-counted. ovulation ≈ next predicted period − 14. The luteal phase is relatively fixed (~10-16 days); the follicular phase is what varies. This is why long cycles delay ovulation but keep the luteal length.
5. Classify against FIGO ranges (§Classifying), not against 28. 24-38 day cycles are normal frequency; irregular is a description, not a defect. PCOS and perimenopause produce genuinely long cycles.
6. Flag against HER baseline once >=3 cycles set it. A sustained shift of >9 days in cycle length, or any Red Flag signal, is worth surfacing. A single off cycle is noise, not a flag.
7. On hormonal contraception, bleeding is not a cycle. Withdrawal and breakthrough bleeds are method effects, not ovulation — never run rules 2-4 on them (contraception.md).

Prediction Math (worked)

Logged Day 1 dates: Jan 3, Jan 31, Feb 26, Mar 27. Cycle lengths: 28, 26, 29 days.

• Median = 28. Shortest 26, longest 29 → half-spread = (29−26)/2 = 1.5, round to 2.
• Next period = Mar 27 + 28 = Apr 24, plus or minus 2 days.
• Ovulation = Apr 24 − 14 = ~Apr 10. Fertile window = Apr 5 through Apr 10.

If she has only 2 cycles, report "roughly late April, still learning your pattern" and refuse a hard date. Recompute the median every new cycle. The window is the last prediction_window cycles (default 6) and never includes cycles older than 12 months — storage keeps everything (her data), but stale drift never re-enters the estimate.

Classifying A Cycle (FIGO thresholds)

These are adult ranges; adolescent and perimenopausal normals differ (irregular.md). Irregular by these axes is common with PCOS, perimenopause, thyroid issues, or high stress. Report the classification; do not name the cause.

Fertility Signals (opt-in only)

Enable only when fertility_tracking is set or she explicitly asks. Three confirming signals, ranked by predictive value:

• LH surge (ovulation predictor kit): positive means ovulation in ~24-36 hours. Best forward predictor.
• Cervical mucus: clear, stretchy, egg-white texture marks peak fertility (the 1-2 days before ovulation).
• Basal body temperature: a sustained rise of 0.3-0.5 C confirms ovulation already happened — retrospective, so it verifies a cycle but cannot predict the current one. Cross-check the calendar estimate (rule 4) against these; when they disagree, live signals win. Protocols, conceive/avoid modes, and test timing: fertility.md.

Red Flags

Anything in this table suspends prediction and coaching: surface it plainly and route to a clinician.

Output Gates

Before answering a cycle question, verify:

• Day 1 anchored on full flow, not spotting?
• Every prediction carries a range computed from her spread, not a bare date?
• What she just reported checked against the Red Flags table?
• Comparing against her baseline, not a textbook 28?
• If contraception is a hormonal method, did contraception.md rules replace ovulation math?
• Mood or symptoms not attributed to her cycle unless she drew the link first?
• Fertility content gated on fertility_tracking or an explicit ask?

Traps

Where Experts Disagree

• Cycle syncing (training/diet by phase). Popular, but generic phase plans rest on weak evidence. Default: her own logged patterns beat any template; offer phase framing only if her data shows the pattern.
• Fertile window width. Calendar-only estimates are the weakest; symptothermal confirmation narrows honestly. Default: calendar for orientation, live signals for decisions — and say which one an answer is based on.
• BBT in the wearable era. Manual BBT is fussy but interpretable; wearable temperature detects the shift with opaque algorithms. Default: treat wearables as BBT-equivalent confirmation, never as forward prediction.

Related Skills

More Clawic skills, get them at https://clawic.com/skills/period (install if the user confirms):

• symptoms — general (non-cycle) symptom tracking and doctor-visit prep
• pregnancy — once a pregnancy is confirmed and tracking shifts to prenatal
• doctor — turn flagged signals into a structured question list for an appointment
• health — longitudinal wellness habits beyond the cycle

Feedback

• If useful, star it: https://clawic.com/skills/period
• Latest version: https://clawic.com/skills/period

Part of Clawic, the verified skill library. Get this skill: https://clawic.com/skills/period.