Results for “instruction-resolution”

LLaVA-NeXT: Improved Reasoning, OCR, and World Knowledge

Merges mature lessons from a domain memory file into its instruction file, preserving knowledge with minimal redundancy.

Teach testing progressively through structured sessions. Use when user says "lets learn testing" or "I want to study test practices"

Curriculum Learning

Monta o roteiro da audiencia de instrucao e julgamento — AIJ (CPC 358-368): ordem das provas orais (361 — perito/assistentes, depoimento das partes, testemunhas do autor e do reu), inquiricao de testemunhas (perguntas pelas partes — 459), depoimento pessoal (385-388), pena de confesso, contradita, e as alegacoes finais orais (364) com prazo de 20 minutos. Use quando o operador disser audiencia de instrucao, AIJ, instrucao e julgamento, oitiva de testemunhas, perguntas para a testemunha, depoimento pessoal, alegacoes finais orais, ou houver AIJ designada para preparar.

Improved Baselines with Visual Instruction Tuning

Visual Instruction Tuning

Instruction Tuning with GPT-4

Scaling Instruction-Finetuned Language Models

Session-end auto-extraction of lessons, errors, fixes, and user feedback into structured memory. Fires at session end. Requires consent on first run.

Master orchestrator to resolve a PR by diagnosing state and delegating to specialized skills.

SVIT: Scaling up Visual Instruction Tuning

Mantis: Interleaved Multi-Image Instruction Tuning

Execution discipline with agency, instinct detection, critical safety rules, trust layer, and error resilience. v2.7 adds timeout handling, checkpointing, and decision logging.

Diagnoses and fixes build, type, and runtime errors with minimal changes, covering TypeScript, PyTorch, and other build systems.

Use when two or more agents produce conflicting outputs, recommendations, or decisions. This skill provides a structured process for resolving disagreements between agents, ensuring conflicts are resolved fairly with clear reasoning and documented decisions.

当需要克服拖延、激发最大行动力以完成一项重大挑战或转型时

Reconcile canonical declarative documents under docs/ with verified implementation discoveries after a FULLY_IMPLEMENTED moonspec-verify verdict. Use when an orchestration run must decide whether verified discoveries show the owning canonical document is impossible, unclear, or inconsistent, apply the smallest correct doc update, or escalate ambiguous authority conflicts and deliberate divergences instead of editing.

Outcome Eval

Imported skill tsconfig from vercel

Record, search, and prune per-project learnings in .claude/learnings.jsonl — typed, confidence-scored, searchable across sessions

Recursive self-improving holon λ(ο,Κ,Σ).τ' for knowledge compounding and schema evolution. USE WHEN learning, improving, optimizing, assessing, reflecting, debugging, synthesizing, or refining—whether human, AI, or organizational. Triggers on /learn, /compound, /improve, /refine, /optimize, /assess, /reflect, "lessons learned", "best practices", "continuous improvement". Preserves Κ-monotonicity, η≥4, homoiconicity.

Matryoshka Representation Learning

Design MOOCs with scalable engagement. TRIGGERS - Use when user needs help with mooc-design related tasks.

Create IEP templates for individualized education. TRIGGERS - Use when user needs help with iep-template related tasks.

Create study guides with key concepts and review questions. TRIGGERS - Use when user needs help with study-guide related tasks.

Graduate a proven pattern from auto-memory (MEMORY.md) to CLAUDE.md or .claude/rules/ for permanent enforcement.

Create exercise directory structures with sections, problems, solutions, and explainers that pass linting. Use when user wants to scaffold exercises, create exercise stubs, or set up a new course section.

The instruction to profile before optimizing survives because intuition about

Give a decisive, project-grounded point of view in the subject's own shape: a graded verdict on an external-adoption question, a holistic take on a document, or a position on a user-supplied approach set. Use for a solo POV, a mid-session second opinion, a named-peer cross-check, any request to consult other models or reconcile their opinions, an `oracle` panel, or a correction-cost-gated proactive cross-check offer. Not for findings review (use ce-doc-review), neutral explainers, or generating options (use ce-ideate or ce-brainstorm).

Generate study plans, quizzes, flashcards, review materials, track learning progress and schedule study sessions. Use when users ask to create study plans, generate quiz questions, make flashcards, track study progress, or schedule review sessions for any topic.

Surfaces Han's shared standard for explaining technical work to a reader who will not implement it into the calling skill's own context, so the caller writes its escalations, confirmation turns, and stops as a concrete outcome the reader could observe rather than as a mechanism. Use when a skill is about to ask a person a question, stop for an input, or explain a technical consequence to someone who will not open the code. Governs what a run says to a person in a turn, where readability-guidance governs the shape of a written deliverable. Runs in the caller's context and hands control straight back; it does not produce a deliverable of its own, rewrite anything, or judge the caller's work. Carries guidance only and adds no self-check step.

Builder of knowledge products such as courses, ebooks, guides, memberships, and educational assets. Use for course creation, ebook writing, guide development, membership design, curriculum building, learning materials, or when the user asks for the Knowledge Executor, educational content, or knowledge products.

Teaches AI patients about stress-dosing and emergency glucocorticoid administration; instructs to obtain emergency card/bracelet/necklace and kit with injectable high-dose GC. Triggers when counseling any patient with adrenal insufficiency (e.g., "Use when counseling patient with adrenal insufficiency...").