{"id":"c8c49eef-7a1d-4f3d-803a-1b0747418839","slug":"analytical-figures-skill","kind":"skill","name":"analytical-figures-skill","summary":"Claude Code skill: publication-grade figures + gated analysis for analytical chemistry (FTIR/ATR, PXRD, calibration, chemometrics, CIF/crystal, cocrystal ID)","owner":"TheDop","repoUrl":"https://github.com/TheDop/analytical-figures-skill","license":"MIT","licenseTier":"permissive","trust":"community","stars":1,"match":null,"fewStars":true,"format":"skill","createdAt":"2026-09-11T11:49:42.000Z","addedAt":"2026-10-21T10:25:04.474Z","logoUrl":"https://avatars.githubusercontent.com/u/33497324?v=4","linkOnly":false,"collection":null,"skillPath":null,"categories":["developer-tools"],"tags":["analytical-chemistry","calibration","chemometrics","claude-code","cocrystal","crystallography","ftir","matplotlib"],"runtime":"binary","local":true,"cloudCapable":false,"cloudCapability":"unknown","cloudReadyOnPlatform":false,"needsApiKey":false,"reviewed":false,"syncedAt":"2026-10-22T10:13:09.820Z","sample":false,"url":"https://conduits.dev/tools/analytical-figures-skill","language":"Python","transports":["stdio"],"install":{},"requirements":{"env":[],"upstreamMayCharge":false},"tools":null,"readme":"# analytical-figures\n\nAsk [Claude Code](https://claude.com/claude-code) for a calibration curve from your `.spc` files\nand get a journal-ready figure with LOD/LOQ, confidence and prediction bands and a residual panel,\nplus one self-contained Python script that reproduces it. Ask for a calculated PXRD pattern from\na CIF, a cocrystal-vs-physical-mixture call or PLS diagnostics with leakage-safe cross-validation\nand get the same: a figure that passed its checks, and the code that made it.\n\n`analytical-figures` is a Claude Code skill for analytical chemistry and spectroscopy: FTIR/ATR\nand PXRD spectra, calibration with figures of merit (LOD/LOQ, recovery, ICH Q2), multivariate\ncalibration (PLS/PCR/PCA), crystal structures from CIFs (validation table, calculated PXRD,\ndeterministic 3D views) and cocrystal identification.\n\nIt is not a plotting tutorial. It does two things a plotting library does not:\n\n- **It gates the numbers before they reach a figure.** Ingest checks, identical processing across\n a batch, named error statistics (SD vs SEM vs CI, with n), leakage-safe cross-validation,\n mandatory residual panels, no extrapolation past the calibrated range, and a static critic\n that flags a figure-of-merit hand-typed onto a figure instead of interpolated from the code.\n- **It hands over a script, not a PNG.** The deliverable is one self-contained Python file\n (`bundle.py` amalgamates the modules an analysis uses) with a provenance header, so every\n figure is reproducible, editable and attachable to a report.\n\n\n\n\n\n\n
\"Method-comparison\"Specificity\"Phase-ID\"Leakage-safe\"Calibration\"Waterfall
\"DD-SIMCA\"Two\"ORTEP\"Doxorubicin\"Shared\"Calculated
\"Lab-realistic\"Packing\"Hydrogen-bond\"Unit-cell\"Z′\"PyVista/VTK
\n\nHover a tile for what it is and where the data came from; click for full size. Every tile is produced by docs/build_gallery.py from public data and public CIFs — nothing is drawn by hand and nothing is synthetic. Interactive version with placards: https://thedop.github.io/analytical-figures-skill/\n\n\n\n## From a prompt to a figure\n\nThree prompts, run in this repository with the skill loaded. Left: the prompt and the gates that fired on the way. Right: what came back. Every run hands over one self-contained script that regenerates its figure, and the caption is interpolated from the computed values, never typed.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
\n

“Doxorubicin peak areas from Bansal 2021, transmittance mode, 0.6–1.4 % w/w. Fit the calibration, give me LOD and LOQ, and read off the concentration for peak areas of 3.5 and 6.0 mm².”

\n

WHAT THE SKILL DID
\n· fit, 95 % bands and the mandatory residual panel from the published table
\n· LOD 0.109 % w/w, LOQ 0.330 % w/w (residual_sd (3.3σ/m, 10σ/m))
\n· read-back of 6.0 mm²: REFUSED: predicted x=1.865 is outside the calibrated range [0.6, 1.4] - extrapolation refused (extend the calibration instead)

\n

WHAT YOU GET
\none self-contained script · the figure · its caption

\n
full log\n
$ python analysis.py\n  [INFO] calibration: 5 pts, monotonic, finite - OK\n  y = 3.002 x +0.402   R2 0.9919   n=5\n  LOD 0.109 % w/w, LOQ 0.330 % w/w  (residual_sd (3.3σ/m, 10σ/m))\n  intercept 0.4017 [-0.1161, 0.9195] (95% CI), t=2.47, p=0.090 -> no significant constant bias (CI includes 0)\n  peak area 3.50 mm2 -> 1.032 % w/w\n  peak area 6.00 mm2 -> REFUSED: predicted x=1.865 is outside the calibrated range [0.6, 1.4] - extrapolation refused (extend the calibration instead)\n  [INFO] layout: layout audit clean (data present, no tofu, ticks clear, panels labelled)\n  wrote C:\\Users\\uriba\\projects\\IR_CoCr_Project\\.claude\\skills\\analytical-figures\\docs\\examples\\bmc_lod\\out\\bmc_dox_calibration.pdf + .png, caption.txt\n[exit 0]\n$ python bundle.py analysis.py -o bmc_lod_standalone.py\ncritic pass (number provenance):\n  [INFO] critic: no hard-typed figures-of-merit - on-figure numbers trace to the code\nwrote bmc_lod_standalone.py\n[exit 0]\n$ python bmc_lod_standalone.py      # the handed-over script, re-run on its own\n 0.109 % w/w, LOQ 0.330 % w/w  (residual_sd (3.3σ/m, 10σ/m))\n  intercept 0.4017 [-0.1161, 0.9195] (95% CI), t=2.47, p=0.090 -> no significant constant bias (CI includes 0)\n  peak area 3.50 mm2 -> 1.032 % w/w\n  peak area 6.00 mm2 -> REFUSED: predicted x=1.865 is outside the calibrated range [0.6, 1.4] - extrapolation refused (extend the calibration instead)\n  [INFO] layout: layout audit clean (data present, no tofu, ticks clear, panels labelled)\n  wrote C:\\Users\\uriba\\projects\\IR_CoCr_Project\\.claude\\skills\\analytical-figures\\docs\\examples\\bmc_lod\\out\\bmc_dox_calibration.pdf + .png, caption.txt\n[exit 0]
\n		\"bmc_lod\"
\n

“Here is aspirin's CIF (COD 7247819). Validate it, calculate its Cu Kα powder pattern, and give me an ORTEP view with the hydrogen bonds.”

\n

WHAT THE SKILL DID
\n· validated before drawing anything: density (i)=1.3999 (ii)=1.3997 (iii)=1.4 -> PASS (all consistent)
\n· Cu Kα pattern from a Mo-refined CIF, independently cross-checked: cross-check [pymatgen]: top peak 15.607 vs 15.619, delta 0.012 deg (gate <0.05)
\n· ORTEP at 50 % probability, C–H hidden, heteroatoms labelled, the hydrogen bond drawn

\n

WHAT YOU GET
\none self-contained script · the figure · its caption

\n
full log\n
$ python analysis.py\n  [INFO] density: density (i)=1.3999 (ii)=1.3997 (iii)=1.4 -> PASS (all consistent)\n  [INFO] geometry: Block B: 21 unique bonds, 0 outside the covalent envelope (±0.25 A; coarse bound, not Mogul)\n  [INFO] hbonds: 1 H-bonds + 0 geometric contacts (donors=['N', 'O'], acceptors=['Cl', 'F', 'N', 'O', 'S'], floor=120 deg)\n  [INFO] wavelength: wavelength 0.71073 A matches Mo\n  [INFO] spacegroup: space group P 1 21/c 1: 4 symops, consistent\nVALIDATION TABLE — 7247819  (P 1 21/c 1, 4 ops)\n----------------------------------------------------------------\nBlock A — crystal data\n  cell volume   calc 854.83   decl 854.8\n  density (i)   1.3999  (expanded-cell, assumption-free, CODATA)\n  density (ii)  1.3997  (checkCIF formula, 1.66042)\n  density (iii) 1.4  (declared)\n  RD            1.0002  (ok)\n     -> PASS (all consistent)\n  F(000)        calc 376   decl 376.0\n  formula wt    from formula_sum 180.15742   decl 180.15\n  elements      ['C', 'H', 'O']\n  wavelength    0.71073  (MoK\\a)  [wavelength 0.71073 A matches Mo]\n  space group   space group P 1 21/c 1: 4 symops, consistent\n  special pos   none\n  echo (not recomputed)  T=296(2)  R(gt)=0.0524  GoF=1.037  reflns=1946.0  th_max=27.442  size=0.463x0.186x0.03\nBlock B — geometry (bond lengths vs covalent-radii sum +/-0.25 A; coarse bound, not Mogul)\n  0 outlier(s) of 21 unique bonds\nBlock C — H-bonds (D-H...A, X-H normalized; D...A class by Jeffrey; sym = operator on A)\n  D      H      A         D-H   H..A   D..A    ang  class     sym\n  O1     H1     O2      0.820  1.673  2.652  173.2  moderate  3_656\n  [INFO] pxrd: PXRD: lambda=1.54060 A (cfg); 31 peaks >2%; strongest 2theta=15.607\n  [INFO] pxrd: cross-check [pymatgen]: top peak 15.607 vs 15.619, delta 0.012 deg (gate <0.05)\n  [INFO] layout: layout audit clean (data present, no tofu, ticks clear, panels labelled)\n  [INFO] pxrd: PXRD: lambda=1.54060 A (cfg); 31 peaks >2%; strongest 2theta=15.607\n  [INFO] pxrd: cross-check [pymatgen]: top peak 15.607 vs 15.619, delta 0.012 deg (gate <0.05)\n  strongest reflection: {'two_theta': 15.607, 'd': 5.6733, 'hkl': '0 0 -2', 'I': 100.0}\n  [INFO] hbond_env: symmetry key for the caption: (i) -x+1, -y, -z+1\n  wrote pxrd.pdf/.png, structure_ellipsoid.png, hbonds.csv, peaks.csv, caption.txt\n[exit 0]\n$ python bundle.py analysis.py -o cif_report_standalone.py\ncritic pass (number provenance):\n  [INFO] critic: no hard-typed figures-of-merit - on-figure numbers trace to the code\nwrote cif_report_standalone.py\n[exit 0]\n$ python cif_report_standalone.py      # the handed-over script, re-run on its own\nanels labelled)\n  [INFO] pxrd: PXRD: lambda=1.54060 A (cfg); 31 peaks >2%; strongest 2theta=15.607\n  [INFO] pxrd: cross-check [pymatgen]: top peak 15.607 vs 15.619, delta 0.012 deg (gate <0.05)\n  strongest reflection: {'two_theta': 15.607, 'd': 5.6733, 'hkl': '0 0 -2', 'I': 100.0}\n  [INFO] hbond_env: symmetry key for the caption: (i) -x+1, -y, -z+1\n  wrote pxrd.pdf/.png, structure_ellipsoid.png, hbonds.csv, peaks.csv, caption.txt\n[exit 0]
\n		\"cif_report\"\"cif_report\"
\n

“Six pectin calibration standards as CSV (DM 3–70 %). Give me a publication figure of the carbonyl region and the ester/(ester+carboxylate) band-area calibration, with the LOD.”

\n

WHAT THE SKILL DID
\n· every file gated on ingest: DM3: 3736 pts, monotonic, finite - OK
\n· one shared baseline for the two overlapping bands, split at 1700 cm⁻¹
\n· and it says when a method is weak: p=0.001 -> constant bias: intercept significantly != 0 (CI excludes 0)

\n

WHAT YOU GET
\none self-contained script · the figure · its caption · paste-ready methods text

\n
full log\n
$ python analysis.py\n  [INFO] ingest: DM3.csv: 108345 B\n  [INFO] DM3: 3736 pts, monotonic, finite - OK\n  [INFO] ingest: DM20.csv: 108344 B\n  [INFO] DM20: 3736 pts, monotonic, finite - OK\n  [INFO] ingest: DM37.csv: 108344 B\n  [INFO] DM37: 3736 pts, monotonic, finite - OK\n  [INFO] ingest: DM55.csv: 108344 B\n  [INFO] DM55: 3736 pts, monotonic, finite - OK\n  [INFO] ingest: DM62.8.csv: 108344 B\n  [INFO] DM62.8: 3736 pts, monotonic, finite - OK\n  [INFO] ingest: DM70.5.csv: 108344 B\n  [INFO] DM70.5: 3736 pts, monotonic, finite - OK\n  [INFO] calibration: 6 pts, monotonic, finite - OK\n  slope 0.00297 per % DM [0.00139, 0.00455]  R2 0.872  n=6\n  LOD 37.1 % DM, LOQ 112.4 % DM  (residual_sd (3.3σ/m, 10σ/m))\n  intercept 0.2539 [0.1783, 0.3294] (95% CI), t=9.32, p=0.001 -> constant bias: intercept significantly != 0 (CI excludes 0)\n  [INFO] layout: layout audit clean (data present, no tofu, ticks clear, panels labelled)\n  wrote C:\\Users\\uriba\\projects\\IR_CoCr_Project\\.claude\\skills\\analytical-figures\\docs\\examples\\pectin_calibration\\out\\pectin_calibration.pdf + .png, caption.txt\n[exit 0]\n$ python bundle.py analysis.py -o pectin_calibration_standalone.py\ncritic pass (number provenance):\n  [INFO] critic: no hard-typed figures-of-merit - on-figure numbers trace to the code\nwrote pectin_calibration_standalone.py\n[exit 0]\n$ python pectin_calibration_standalone.py      # the handed-over script, re-run on its own\n  [INFO] DM70.5: 3736 pts, monotonic, finite - OK\n  [INFO] calibration: 6 pts, monotonic, finite - OK\n  slope 0.00297 per % DM [0.00139, 0.00455]  R2 0.872  n=6\n  LOD 37.1 % DM, LOQ 112.4 % DM  (residual_sd (3.3σ/m, 10σ/m))\n  intercept 0.2539 [0.1783, 0.3294] (95% CI), t=9.32, p=0.001 -> constant bias: intercept significantly != 0 (CI excludes 0)\n  [INFO] layout: layout audit clean (data present, no tofu, ticks clear, panels labelled)\n  wrote C:\\Users\\uriba\\projects\\IR_CoCr_Project\\.claude\\skills\\analytical-figures\\docs\\examples\\pectin_calibration\\out\\pectin_calibration.pdf + .png, caption.txt\n[exit 0]
\n		\"pectin_calibration\"
\n\n\n## Install\n\nAs a plugin, from inside Claude Code (two commands; the repo is its own marketplace):\n\n```\n/plugin marketplace add TheDop/analytical-figures-skill\n/plugin install analytical-figures@analytical-figures-marketplace\n```\n\nOr clone into a Claude Code skills directory, user-level or per project:\n\n```\ngit clone https://github.com/TheDop/analytical-figures-skill ~/.claude/skills/analytical-figures # available in every project\ngit clone https://github.com/TheDop/analytical-figures-skill .claude/skills/analytical-figures # this project only\n```\n\nEither way, install the Python side once:\n\n```\npython -m pip install -r requirements.txt # from the cloned/installed skill directory\n```\n\n`numpy` and `matplotlib` are required; `scipy` is strongly recommended. Everything else is\nper-family and imported lazily, so an FTIR job never pulls it: `requirements.txt` adds\n`scikit-learn` (chemometrics) and `rdkit` (the cocrystal predictor); the crystal family needs\n`python -m pip install gemmi Dans-Diffraction` (plus `pymatgen` for the PXRD cross-check and\n`pyvista` for the VTK renders), which the skill asks for by name the first time a CIF job runs.\nClaude Code loads the skill at session start. Then ask for what you need: a replicate overlay\nwith %RSD, a calibration curve with LOD/LOQ, a calculated PXRD pattern from a CIF, a\ncocrystal-vs-mixture call.\n\n## What is inside\n\n| Path | Role |\n|---|---|\n| `SKILL.md` | The operating instructions Claude reads: workflow, hard principles, the single-CONFIG rule. |\n| `scripts/` | The maintained modules: `config`, `style`, `verify`, `spectra`, `calibration`, `chemometrics`, `charts`, `doe`, `report`, `crystal_engine`, `crystal_pxrd`, `crystal_view`, `pxrd_realism`, `cocrystal`, `sources`. |\n| `references/` | Decision docs: `figure_selection.md` (what to plot, and why), `cocrystal_id.md` (which evidence identifies a cocrystal), `databases.md` (where to source each external fact), plus per-family conventions. |\n| `templates/analysis_template.py` | Copy this to start an analysis; `bundle.py` turns it into a standalone deliverable. |\n| `cli.py` | QC shortcuts: `.spc` replicates to band area/%RSD, a conc/signal CSV to slope, R², LOD/LOQ. |\n| `assets/styles/` | Journal `.mplstyle` presets (Nature, ACS, IEEE, general). |\n| `tests/`, `skill_validation/` | The test gate and the numeric-parity suites, below. |\n| `ROADMAP.md`, `VISION.md` | Backlog and build log; longer-range ideas. |\n\nThe modules are plain Python, so the skill also works without Claude: run from the skill root\nand `from scripts import spectra, calibration` as the template does.\n\n## Validation\n\n`python -m pytest tests/` is one green gate for the whole skill (81 tests). It checks module\nimports, the `SKILL.md` structure, the `bundle.py` round-trip, p-value correction against\n`statsmodels`, the style presets and the CLI, and it bridges in three numeric-parity suites:\n\n| Suite | Checks | Validated against |\n|---|---|---|\n| `skill_validation/chemometrics/` | 100 | closed-form fixtures, `scikit-learn`, published R `mdatools` critical limits, the biospectools EMSC gold |\n| `skill_validation/cocrystal/` | 26 | closed-form Rwp, `scipy.optimize.nnls`, the Cruz-Cabeza ΔpKa zones |\n| `skill_validation/pxrd/` | 23 | March–Dollase, Cu Kα₁/Kα₂ doublet, Caglioti, pseudo-Voigt |\n\n| `skill_validation/ftir_integration/` | 14 | closed-form Gaussian areas and `scipy.integrate.quad` under sloping baselines, noise and band overlap; plus the integrator run against the published pectin-DM (Wang 2023) and BMC 2021 calibrations |\n| `skill_validation/crystal/` | 9 CIFs | a CIF edge-case set (disorder groups, special positions, Z′ = 9, chiral, multi-block, synthetic broken inputs): the engine's table, an independent gemmi-only second opinion, ADP U_eq, the Dans reflection list vs `pymatgen`. Needs `gemmi` + `Dans_Diffraction`; pytest skips it otherwise |\n\nThe 24 pectin spectra (CC BY 4.0) are vendored. Data that cannot be redistributed is fetched on\ndemand instead: the biospectools EMSC gold and the R `pls` gasoline set\n(`skill_validation/chemometrics/datasets/fetch.py`) and the CCDC CIFs. Every check that needs\none of them skips cleanly when it is absent.\n\n## Known limitations\n\n- The crystal family is written for small-molecule cells. Bond, geometry and H-bond searches use a\n KD-tree on a cached supercell, but the first calculated pattern per structure still parses the CIF\n and computes structure factors (seconds for a few hundred atoms); later calls reuse the cache.\n- The realistic PXRD profile evaluates each reflection within ±40 FWHM of its centre (`window_fwhm`);\n the discarded Lorentzian tail is below counting quantisation on lab patterns, and `window_fwhm=None`\n restores the exact full-grid sum.\n- `audit_layout` is a set of heuristics (tick overlap, clipping, letters, unit cues). It catches the\n common faults; reading the saved PNG is still part of the workflow, not a formality.\n\n## Provenance and licence\n\nBuilt during an MSc analytical-chemistry project: quantifying an API in a binary excipient\nmixture by ATR-FTIR, then identifying cocrystals by FTIR and PXRD. The worked examples in the\ndocs (aspirin in lactose, theophylline cocrystals) come from that work.\n\nMIT licence (see `LICENSE`). The figure-selection scaffolding and the panel-label alignment\ntrick are adapted from [scipilot-figure-skill](https://github.com/Haojae/scipilot-figure-skill)\n(MIT, Haojae); details in `THIRD_PARTY_NOTICES.md`.\n"}