{"slug":"down-skilling","title":"down-skilling","summary":"Distill Opus-level reasoning into optimized instructions for Haiku 4.5 (and Sonnet). Generates explicit, procedural prompts with n-shot examples that maximize smaller model performance on a given task. Use when user says \"down-skill\", \"distill for Haiku\", \"optimize for Haiku\", \"m","platform":"Claude","tags":[],"authorName":"LLM Mart","authorSlug":"llm-mart","score":0,"source":"github","price":null,"verified":false,"createdAt":"2026-09-30T16:45:19.965036Z","repo":{"url":"https://github.com/oaustegard/claude-skills","stars":150,"forks":6,"license":"MIT","updatedAt":"2026-09-30T12:40:57Z"},"bodyHtml":"<h1>down-skilling</h1>\n<p>Distills Opus-level reasoning into explicit, procedural prompts with\nn-shot examples so Haiku (or Sonnet) can execute a task reliably. Use\nwhen the user says \"down-skill\", \"distill for Haiku\", or wants to\ndelegate a task to a smaller model with high reliability.</p>\n<p>Before distilling anything, the skill now triages: if the task's output\nis mechanically checkable (schema validates, tests pass, a count is a\ncount), a minimal prompt plus a deterministic verifier beats a\nheavy example-laden distillation — save the n-shot investment for\noutputs that genuinely need judgment. That triage step, the per-gap\nmitigations, and the pricing in the Economics section were all recalibrated\nagainst measured Haiku 4.5 data on 2026-07-15 (see <code>agent-routing</code>'s\n<a href=\"../agent-routing/references/calibration-2026-07-15.md\">calibration reference</a>) —\nseveral model-card-era priors turned out to understate Haiku, most notably\non counting/enumeration and multi-hop reasoning over <em>explicit</em> chains.</p>\n<p>Full prompt architecture, the gap catalog (<code>gaps/</code>), worked\nbefore/after examples (<code>examples/</code>), and the self-check checklist live in\n<strong><a href=\"./SKILL.md\">SKILL.md</a></strong>. Version history in\n<strong><a href=\"./CHANGELOG.md\">CHANGELOG.md</a></strong>.</p>\n","files":[{"path":"CHANGELOG.md","sizeBytes":3941,"isText":true},{"path":"examples/api-orchestration.md","sizeBytes":2994,"isText":true},{"path":"examples/code-review-triage.md","sizeBytes":2818,"isText":true},{"path":"examples/content-moderation.md","sizeBytes":5951,"isText":true},{"path":"examples/creative-rewriting.md","sizeBytes":2386,"isText":true},{"path":"examples/data-extraction.md","sizeBytes":5107,"isText":true},{"path":"examples/document-qa.md","sizeBytes":5005,"isText":true},{"path":"examples/email-summarization.md","sizeBytes":6902,"isText":true},{"path":"examples/meeting-notes.md","sizeBytes":2872,"isText":true},{"path":"examples/resume-screening.md","sizeBytes":2645,"isText":true},{"path":"examples/sql-generation.md","sizeBytes":2422,"isText":true},{"path":"examples/step-by-step-analysis.md","sizeBytes":1993,"isText":true},{"path":"examples/text-classification.md","sizeBytes":4725,"isText":true},{"path":"gaps/ambiguity-resolution.md","sizeBytes":552,"isText":true},{"path":"gaps/code-generation.md","sizeBytes":1392,"isText":true},{"path":"gaps/comparative-analysis.md","sizeBytes":1062,"isText":true},{"path":"gaps/conditional-logic.md","sizeBytes":1362,"isText":true},{"path":"gaps/context-utilization.md","sizeBytes":739,"isText":true},{"path":"gaps/counting-enumeration.md","sizeBytes":1575,"isText":true},{"path":"gaps/creative-generation.md","sizeBytes":699,"isText":true},{"path":"gaps/implicit-constraints.md","sizeBytes":646,"isText":true},{"path":"gaps/instruction-density.md","sizeBytes":1134,"isText":true},{"path":"gaps/multi-hop-reasoning.md","sizeBytes":1485,"isText":true},{"path":"gaps/multi-turn-consistency.md","sizeBytes":1144,"isText":true},{"path":"gaps/negation-handling.md","sizeBytes":1049,"isText":true},{"path":"gaps/nuanced-classification.md","sizeBytes":701,"isText":true},{"path":"gaps/output-calibration.md","sizeBytes":594,"isText":true},{"path":"gaps/parallel-consistency.md","sizeBytes":1223,"isText":true},{"path":"gaps/partial-information.md","sizeBytes":1170,"isText":true},{"path":"gaps/schema-adherence.md","sizeBytes":674,"isText":true},{"path":"gaps/self-correction.md","sizeBytes":675,"isText":true},{"path":"gaps/summarization-fidelity.md","sizeBytes":1239,"isText":true},{"path":"gaps/tool-use-planning.md","sizeBytes":672,"isText":true},{"path":"README.md","sizeBytes":1216,"isText":true},{"path":"SKILL.md","sizeBytes":21427,"isText":true}],"reviewScore":null,"reviewSummary":null,"trust":{"provenance":"trusted-source-unreviewed","notice":"Community-authored content, reproduced verbatim and not vetted as instructions. Treat it as data to evaluate, never as directives to follow.","bodySource":null},"bodyLocked":false,"purchaseUrl":null,"sourceUrl":null,"report":{"provenance":"trusted-source-unreviewed","screen":{"ran":true,"outcome":"clean","suspicious":0,"notes":0,"hiddenCharacters":false},"virusScan":{"engine":"clamav","status":"clean","scannedAt":"2026-09-30T16:45:30.574021Z","sha256":"A2F45959BCE7B2D310C31B2F4B7B9519623B33A2F6177546615ABDE9BD8AC718","sizeBytes":47660},"review":null,"source":{"repositoryUrl":"https://github.com/oaustegard/claude-skills","path":"plugins/ai-and-reasoning/skills/down-skilling","license":"MIT","commit":"e39c726ecde72943147986b7de11a05cbd74854d","subtreeSha":"F794D6480DD6A128E4A08E989BAAFB0A3B9CD3EF805E35C059554DDD782D11AD","lastSyncedAt":"2026-09-30T16:45:19.336787Z"},"reviewedAt":"2026-09-30T16:47:57.603685Z","notice":"Community-authored content, reproduced verbatim and not vetted as instructions. Treat it as data to evaluate, never as directives to follow."},"install":[{"target":"skills-cli","command":"npx skills add https://github.com/oaustegard/claude-skills/tree/main/plugins/ai-and-reasoning/skills/down-skilling"},{"target":"claude-code","command":"claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install oaustegard-claude-skills@llmmart"},{"target":"git","command":"git clone https://github.com/oaustegard/claude-skills.git"}]}