Skip to content

Releases: gronbow/GeoSkills

GeoSkills v0.4.0 — Unified local workflow

Choose a tag to compare

@gronbow gronbow released this 08 Aug 09:50
1778f50

Highlights

  • Added one provider-neutral geoskills.py entry point with version, self-check, review-only plan, and guarded run commands.
  • Added strict geoskills.recipe/v1 YAML recipes and content-addressed plans that become invalid when the recipe, input, tool version, style, task, or scientific asset changes.
  • Added atomic multi-task output for the four reviewed diagram families: REE, trace-element spider, Harker, and volcanic TAS.
  • Added privacy-safe shareable reports and an explicit local-reproducible profile for sensitive plotted-data CSV files.
  • Preserved the reviewed v0.3 single-script command contracts and scientific assets.
  • Preserved editable SVG/PDF and 600 dpi PNG/TIFF publication outputs.

Scientific and privacy boundaries

  • This release adds no new scientific diagram and changes no normalization values or classification boundaries.
  • Scientific calculations remain deterministic and local; plotting scripts do not upload user data.
  • Shareable plans and reports exclude absolute paths, sample identifiers, source values, and exact source-data extrema.
  • Human confirmation remains required for scientific applicability and sensitive-output decisions.

Validation

  • Skill structure validation: passed.
  • Runtime self-check: ready.
  • 196 automated tests passed, including synthetic-data and published-data regression coverage.
  • Post-merge GitHub Actions passed on Ubuntu and Windows with Python 3.11 and 3.12.

See the changelog for the complete release record.

GeoSkills v0.3.0 — REE, spider, Harker, and TAS

Choose a tag to compare

@gronbow gronbow released this 30 Jul 10:51
1fc6ede

GeoSkills v0.3.0 is the first complete public release covering four common geochemical visualization workflows. The reviewed v0.2.0 spider-diagram milestone is included in this release rather than tagged separately.

Workflows

  • Sun and McDonough (1989) C1 chondrite-normalized La–Lu REE patterns
  • primitive-mantle, modified primitive-mantle, or N-MORB normalized trace-element spider diagrams
  • customizable single- or multi-panel Harker variation diagrams
  • guarded IUGS volcanic TAS classification with per-sample review states

Inputs and outputs

  • Read flat or unambiguous transposed CSV, comma/tab-delimited TXT, and XLSX tables.
  • Export editable SVG/PDF, 600 dpi PNG/TIFF, exact plotted-data CSV, and machine-readable JSON reports.
  • Keep private inputs and generated outputs local and outside Git.

Scientific safeguards

  • Require explicit ppm or wt% units; preserve blanks and below-detection-limit values as missing.
  • Keep printed and footnote-modified Sun and McDonough primitive-mantle values separate.
  • Do not add default Harker regressions or infer petrogenetic processes from covariance alone.
  • Require explicit volcanic applicability and composition-basis declarations for TAS.
  • Preserve TAS boundary, ambiguous-field, and out-of-scope cases for review instead of forcing a classification.
  • Do not claim support for isotope or tectonic-discrimination diagrams in this release.

Validation

  • 69 automated tests passed locally.
  • GitHub Actions passed on Ubuntu and Windows with Python 3.11 and 3.12.
  • Harker output was checked with locally held published data; TAS was checked with the committed synthetic fixture.
  • Export QA confirmed embedded PDF fonts, editable SVG text, 600 dpi PNG/TIFF, and LZW TIFF compression.
  • Skill structure validation and Python compilation passed.

Install or upgrade

Copy skills/geoskills to $CODEX_HOME/skills/geoskills. On a default Windows installation, use %USERPROFILE%\.codex\skills\geoskills, then start a new Codex task.

Scientific references

GeoSkills v0.1.0 — REE patterns

Choose a tag to compare

@gronbow gronbow released this 28 Jul 08:46

GeoSkills 的首个公开测试版,只提供球粒陨石标准化 REE 配分图功能。

主要功能:

  • 读取 CSV、TXT 和 Excel 地球化学表格;
  • 使用 Sun & McDonough (1989) C1 球粒陨石值标准化 La-Lu;
  • 输出 SVG、PDF、600 dpi TIFF/PNG、绘图数据和运行报告;
  • 支持完整四边框、自动避让图例和整洁的自适应对数纵轴;
  • 数据全程在本地处理。

验证:23 项自动测试,并在 Windows/Linux、Python 3.11/3.12 上通过 GitHub Actions。

这是功能测试版。正式投稿前,请由研究者核对数据列、单位、标准化方案和最终图件。