Skill Claude CodeCodex
Use when a request needs design tradeoffs for a new capability or a material external behavior, interface, or module change. Explore the design, establish its Spec identity through manage-specs, and stop before implementation.
在 Agent 相关实践中发现,盲目堆砌大量技能并不会提升智能体性能,技能数量无法等价于执行质量。因此我整合了当前业内认可度较高的优质 Skill 资源,搭建本仓库,专供自己科研课题与实验开发使用。
Skill Claude CodeCodex
Use when a request needs design tradeoffs for a new capability or a material external behavior, interface, or module change. Explore the design, establish its Spec identity through manage-specs, and stop before implementation.
Skill Claude CodeCodex
Traceability and convergence audit for a completed Spec Bundle. Use when the user explicitly asks to compare implementation with its Spec/Plan contract or append findings from the just-completed convergence audit as Convergence Tasks.
Skill Claude CodeCodex
A manually triggered check of an answer, plan, or decision for uncertainty and overlooked risks.
Skill Claude CodeCodex
Maintain next-generation documents through four explicit modes. Use when the user wants to check document health, rebuild generated Indexes, update Current Architecture, or recover reviewable documentation after damage or neglect.
Skill Claude CodeCodex
Decomposition of an Accepted Spec and its current Approved Plan into one reviewable, independently executable tasks.md. Use when an existing Spec Bundle needs Tasks created, revised, audited, or approval-recorded before implementation.
Skill Claude CodeCodex
Grill the user relentlessly about a plan, decision, or idea. Use when the user wants to stress-test their thinking, or uses any 'grill' trigger phrases.
Skill Claude CodeCodex
Use when the user explicitly asks to save a session handoff so a fresh agent can continue the work.
Skill Claude CodeCodex
Landing: use only when the user explicitly names landing, or explicitly asks in this turn to pressure-test a recoverable prior direction for real-world feasibility—ground or narrow it, and define a stop rule. A joint takeoff + landing request stays in Takeoff. Enter Landing only after a later explicit request.
Skill Claude CodeCodex
Classify a design request into one stable Spec identity. Use before creating, revising, or replacing a Spec, or when another design Skill needs the canonical Spec owner.
Skill Claude CodeCodex
Use before launching or updating a clearly formal, costly, or retained-evidence research run: a full/baseline/release Benchmark or Eval, full training, a long-running or expensive GPU/remote job, retained checkpoints/predictions/results, evidence for acceptance/publication/product decisions/downstream experiments, or…
Skill Claude CodeCodex
A workflow for rethinking a project's target when the user explicitly asks to broaden the design. It considers the right overall direction instead of limiting the answer to the smallest compatible change.
Skill Claude CodeCodex
Use only when the user explicitly requests TDD, test-first, or Red-Green-Refactor, or the current Approved Task explicitly requires a TDD Process.
Skill Claude CodeCodex
Worktree isolation only when the user explicitly requests a Git worktree, an Approved Task explicitly requires a Worktree Process, or the user explicitly approves an Agent's concrete isolation-risk recommendation.
Skill Claude CodeCodex
Use when starting any conversation - establishes how to find and use hello-scholar skills across all skill groups, requiring skill invocation before ANY response including clarifying questions.
Skill Claude CodeCodex
Reference for writing and editing skills well — the vocabulary and principles that make a skill predictable.
Skill Claude CodeCodex
Create a high-level implementation strategy for an Accepted Spec when the user needs a reviewable implementation Plan. Produces one Bundle plan.md and routes an approved Plan to generating-tasks.