From smp-kat-tools
Surface examiner-flagged failure patterns from RACGP, ACRRM, AMC, and PESCI exam reports. Use when the user asks "what do RACGP examiners flag", "common KFP errors", "AMC clinical exam pitfalls", or wants to map a candidate's wrong answer onto a recognised examiner-criticism class. Returns the wave-1 findings catalogue (16-class AU misconception taxonomy) plus per-college style notes.
How this skill is triggered — by the user, by Claude, or both
Slash command
/smp-kat-tools:examiner-feedback-miningThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Pattern library distilled from Australian college examiner reports. Use this skill before authoring any practice stem, before reviewing a candidate answer, or when the user asks why a model answer differs from the marking key.
Pattern library distilled from Australian college examiner reports. Use this skill before authoring any practice stem, before reviewing a candidate answer, or when the user asks why a model answer differs from the marking key.
Activate when the request involves:
Do NOT activate for: general medical content lookup (use medical-acquisition), single-MCQ distractor authoring (use distractor-design), KFP option-by-option failure routing (use kfp-failure-bucket).
Distilled from RACGP KFP cycles 2018 to 2024 plus AMC clinical exam reports 2019 to 2023.
Examiners reward specificity. A KFP answer that says "antibiotics" when the marking key wants "amoxicillin 500 mg orally three times daily for five days" scores zero. Generic responses fail. Aim for drug name plus dose plus route plus frequency plus duration. For investigations, name the exact test (FBE, UEC, CRP, LFTs) rather than "blood tests".
Examiners punish over-investigation. KFP cycles 2022 and 2023 explicitly flagged candidates who ordered CT before clinical exam, MRI before plain film, and tumour markers in screening contexts.
Multiple-best-answer. Examiners flag candidates who pick a defensible but second-line option when a first-line option is present. Read the stem for context cues (rural, paediatric, pregnancy) that shift the first-line choice.
Communication, structure, safety-netting. Examiners explicitly score "explained the plan in plain language", "offered a follow-up window", "screened for red flags" as separate marking points. Failing one of these is a common path to a marginal pass that becomes a fail.
Resource-constrained context is mandatory. Solutions assuming tertiary access fail. Examiners reward retrieval planning, telehealth referral pathways, and procedural backup arrangements.
Junior-doctor context. Examiners reward clear escalation lines, explicit handover, and recognition of when a case is beyond scope. Empathy and cultural safety are scored separately from the medical decision.
When a user asks "why is my answer wrong", run this sequence:
racgp_exam or acrrm_exam backendWhen authoring a stem, run this sequence:
~/projects/personal/smp-kat-study/knowledge/examiner-feedback-wave1.md (when present)~/tools/bin/acquisition fetch '{"cycle":"2024.2","exam_type":"akt"}' --backend racgp_exampubmed or direct AMC URL through institutional_proxynpx 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.