From smp-kat-tools
Author or evaluate MCQ and KFP distractors against a 72-entry taxonomy plus the 5 KFP failure buckets. Use when the user asks to write a stem with distractors, evaluate whether existing distractors discriminate, classify a wrong-answer pattern, or stress-test an option set for tautology, give-aways, or examiner-flagged sloppiness. Returns taxonomy class, discrimination notes, and a rewrite suggestion when the option fails.
How this skill is triggered — by the user, by Claude, or both
Slash command
/smp-kat-tools:distractor-designThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Operational taxonomy for authoring MCQ / KFP distractors that discriminate. Built from psychometric literature (Haladyna 2016, Case and Swanson NBME guide) plus AU exam style cues from RACGP and AMC reports.
Operational taxonomy for authoring MCQ / KFP distractors that discriminate. Built from psychometric literature (Haladyna 2016, Case and Swanson NBME guide) plus AU exam style cues from RACGP and AMC reports.
Activate when the request involves:
Do NOT activate for: stem-level failure analysis (use kfp-failure-bucket), examiner-style review of an entire answer (use examiner-feedback-mining), retrieval of source material (use medical-acquisition).
Each distractor falls into exactly one category. The author's job is to spread distractors across categories so no two options share a category in the same stem. A stem with two distractors in category 3 is psychometrically wasteful.
1.1 Right indication, wrong class (NSAID for migraine prophylaxis instead of beta-blocker) 1.2 Same class, weaker evidence (atenolol for prophylaxis instead of propranolol) 1.3 Same class, contraindicated subgroup (non-cardioselective beta-blocker in asthma) 1.4 Adjacent class with overlapping receptor (alpha-blocker for migraine instead of beta-blocker) 1.5 Outdated standard (ergotamine where triptan is now first-line) 1.6 Veterinary or off-label adjacent
2.1 Adult dose in paediatric patient 2.2 Loading dose used as maintenance 2.3 Frequency error (BD when TDS required) 2.4 Duration error (5 days when 10 days mandated) 2.5 Route error (oral when IV indicated) 2.6 Renal / hepatic dose-adjustment ignored
3.1 Right modality, wrong tier (CT before X-ray) 3.2 Right test, wrong context (D-dimer in low-pretest probability) 3.3 Outdated marker (CK-MB when troponin is standard) 3.4 Screening test misused as diagnostic 3.5 Specialist test in primary care first-line slot 3.6 Test that does not exclude the sentinel diagnosis
4.1 BP target wrong for population (140/90 when CKD wants 130/80) 4.2 HbA1c target wrong for age / frailty 4.3 Cholesterol target wrong for primary vs secondary prevention 4.4 Anticoagulation target INR wrong for indication 4.5 Threshold for treatment vs threshold for screening conflated 4.6 Population threshold applied to individual
5.1 Right action, wrong window (thrombolysis past 4.5 h) 5.2 Premature intervention (statin before lifestyle trial) 5.3 Delayed intervention (deferring biopsy past red-flag window) 5.4 Wrong follow-up interval (6 months when 6 weeks) 5.5 Pre-op / peri-op timing error 5.6 Pregnancy trimester mismatch
6.1 Adult guideline applied to child 6.2 Non-pregnant guideline applied in pregnancy 6.3 Non-Indigenous guideline applied without ATSI adjustment 6.4 Urban-tertiary plan applied in rural-remote 6.5 Non-immunocompromised plan in immunocompromised 6.6 Adult plan in older frail patient
7.1 Topical when systemic indicated 7.2 Systemic when topical sufficient 7.3 Wrong injection site / depth 7.4 Wrong inhaler device for age 7.5 Oral when patient NBM 7.6 Rectal when oral tolerated
8.1 ED when GP scope sufficient 8.2 GP when ED indicated 8.3 Specialist referral when GP follow-up sufficient 8.4 GP follow-up when specialist mandated 8.5 Inpatient admission when ambulatory care fits 8.6 Discharge when admission criteria met
9.1 Capacity not assessed 9.2 Gillick competence not applied 9.3 Mandatory reporting missed 9.4 Notifiable disease not flagged 9.5 Fitness to drive not addressed 9.6 Advance care directive not respected
10.1 Interpreter not offered 10.2 ATSI cultural framing missing 10.3 Trauma-informed approach missing 10.4 Plain-language explanation missing 10.5 Shared decision making missing 10.6 Safety-netting missing
11.1 GPMP / TCA item number error 11.2 MHCP item number error 11.3 Telehealth item number error 11.4 Care plan content gap 11.5 Recall and reminder system gap 11.6 Privacy or My Health Record handling
12.1 Restates the stem in different words 12.2 Catch-all "all of the above" pattern 12.3 Negative-knowledge probe ("which is NOT") 12.4 Convergent absurdity (clearly wrong, fails as distractor) 12.5 Give-away with grammatical mismatch 12.6 Length give-away (correct option visibly longer)
For KFP short-answer banks, distractors are written-in candidate responses rather than fixed options. The 5 buckets identify the dominant wrong-answer pattern at the option layer.
A. Vague generic ("antibiotics", "blood tests", "scan") that fails the specificity bar B. Defensible but second-line (correct class, wrong first-line choice) C. Out-of-scope escalation or non-escalation D. Drug-dose-frequency-duration partial (drug correct, one component wrong) E. Tautology or restated stem
~/projects/personal/smp-kat-study/npx claudepluginhub anon2023-halmoni/claude-fleet-marketplace --plugin smp-kat-toolsGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.