メニュー
ALWAYS invoke this skill when the user creates, edits, optimizes, benchmarks, or evaluates a Claude Code skill, or asks to improve a skill's activation rate. Do not write or modify SKILL.md files without loading this skill first.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | skill-creator |
| description | ALWAYS invoke this skill when the user creates, edits, optimizes, benchmarks, or evaluates a Claude Code skill, or asks to improve a skill's activation rate. Do not write or modify SKILL.md files without loading this skill first. |
All leaf files in this skill (auto-loaded):
[WHAT] Scitexification — the verb: translate existing code (a script, a notebook, a small repo, a published-paper supplement) into SciTeX form so it gains the ecosystem's session-managed I/O, scitex-clew evidence binding, publication-quality figures, and project structure. Single source of truth for the migration act itself, package-agnostic; per-package patterns delegate to the per-pkg SKILL.md, and Clew-specific specialization layers on top of this skill via `scientific/04_clew_*`. [WHEN] You have existing code that works and want to bring it into the SciTeX ecosystem with minimal rewrite, OR an agent is asked to translate a research capsule, paper supplement, or one-off notebook into a SciTeX project. Load this skill BEFORE picking the per-chapter topic. [HOW] Read this SKILL.md first to orient on the 5-stage translation arc; then drill into the topic chapter you need next (`01_io-patterns`, `02_session-config`, `03_plt-patterns`, `04_repro-clew`, `05_naming-and-numbering`). Per-package details (full stx.i
[WHAT] Scientific-methodology skills for the SciTeX ecosystem — publication-quality figures, statistics, experiment reproducibility. Distinct from `general/` (which covers package engineering) and per-package skills (which cover package-specific APIs). Load when authoring analysis scripts, preparing figures for manuscripts, or checking scientific rigour of ecosystem output. [WHEN] Authoring analysis scripts, preparing figures or PDF reports for manuscripts, or auditing scientific outputs of any ecosystem package. [HOW] Read SKILL.md as the index, then drill into the numbered sub-skill leaves (`01_figures_*`, `02_research-project_*`, `03_reporting_*`) for the relevant topic.
[WHAT] Developer utilities for maintaining the whole SciTeX ecosystem across ~15 packages — version audit and mismatch fixing, editable-install sync, per-host git pull/commit/diff via SSH, bulk rename across files + contents + directories + symlinks (with git-safety guards), aggregated docs search across all packages, a Result/ErrorCode envelope shared by CLI + MCP, DevConfig/HostConfig, a local +… [WHEN] Use whenever the user asks to "list scitex versions", "check for version mismatches", "fix version drift", "rename old_name → new_name across all files", "sync my changes to the HPC / laptop / servers", "run tests on the HPC via Slurm", "poll my sbatch job", "fetch HPC test output", "search the docs for save_fig", "commit remote changes on gpu01", "pull all repos", "audit the ecosystem before a… [HOW] See sub-skills index for entry points.
[WHAT] Canonical engineering standards every `scitex-*` package and research project must follow — 3-layer architecture cascade, dependency/version pinning, local-state directory layout (`~/.scitex/<pkg-short>/` and `<project>/.scitex/<pkg-short>/`), repo layout (src/tests/scripts/examples), the five interfaces (Python API → CLI → MCP → Skills → optional HTTP) with their delegation rules and the noun-verb CLI convention, README/Sphinx docs, version-control workflow plus `scitex-dev ecosystem …` release automation, skill-authoring rules (layout, editable-vs-wheel install, public-vs-private), and the periodic quality checklist. Use as the single entry point for creating, auditing, reviewing, or releasing any SciTeX package. [WHEN] Creating, auditing, reviewing, or releasing any SciTeX package, or onboarding a research repo to ecosystem standards. [HOW] Read SKILL.md as the index, then drill into the per-category sub-dirs (`01_ecosystem/`, `02_package/`, `03_interface/`, `04_docs/`, `05_development/`, `09_qualit
[WHAT] AST-based linter for reproducible-research Python — 47 built-in rules across 7 categories: STX-I (imports), STX-IO (forbid raw pd.read_csv / np.load / pickle / fig.savefig — use scitex.io.save / scitex.io.load instead), STX-P (path handling — resolve via scitex.path, no hardcoded `/home/...`), STX-PA (plot/axes), and more. [WHEN] User asks to lint scitex code, check scitex conventions, list linter rules, show STX-IO001 meaning, or enforce reproducible-research style; mentions STX-*, scitex-dev linter, scitex conventions, reproducibility lint. [HOW] `scitex-dev linter check <path>` (CLI) or `scitex_dev.linter.check(path)` (Python). Hook integration auto-runs on PreToolUse for Edit/Write.
ALWAYS invoke this skill when the user wants to navigate websites, fill forms, take screenshots, test web UIs, extract page data, or debug frontend behavior. Do not spawn browser automation without loading this skill first.
