npm - @monoes/monomindcli - Versions diffs - 2.15.3 → 2.15.5 - Mend

@monoes/monomindcli 2.15.3 → 2.15.5

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Files changed (1403) hide show

package/org-skills/seo-audit/references/international-seo.md ADDED Viewed

@@ -0,0 +1,230 @@
+# International SEO: Evidence & Sources
+Detailed evidence backing the International SEO & Localization section of the SEO Audit skill. Organized by topic with source URLs and key quotes.
+---
+## Hreflang
+### Placement Methods
+Google supports three equivalent methods: HTML `<link>` in `<head>`, HTTP `Link` headers, and XML sitemap `<xhtml:link>` elements. Google confirmed no method is prioritized over another.
+Google combines signals from both HTML and sitemaps. If the same language-region pair points to different URLs across methods, Google drops that pair rather than guessing.
+- [Google Search Central: Localized Versions](https://developers.google.com/search/docs/specialty/international/localized-versions)
+- [SEJ: Google Combines Hreflang Signals](https://www.searchenginejournal.com/google-combines-hreflang-signals-from-html-sitemaps/389219/)
+### Reciprocal Requirement
+Google's docs: "If page X links to page Y, page Y must link back to page X. If not, those annotations may be ignored or not interpreted correctly."
+Every page must include itself (self-referencing) in the hreflang set. Missing self-referencing is the #1 error found by Semrush audits. A study of 374,756 domains found 67% of hreflang implementations had issues.
+- [Google Search Central: Localized Versions](https://developers.google.com/search/docs/specialty/international/localized-versions)
+- [Semrush: 9 Common Hreflang Errors](https://www.semrush.com/blog/hreflang-errors/)
+- [SE Land: 31% of International Websites Contain Hreflang Errors](https://searchengineland.com/study-31-of-international-websites-contain-hreflang-errors-395161)
+### x-default
+Introduced April 2013. Designates the fallback page for users whose language/region matches no declared variant. Can point to the same URL as one of the language-specific alternates. Must be included in the complete set of annotations on every variant page.
+- [Google Blog: x-default hreflang](https://developers.google.com/search/blog/2013/04/x-default-hreflang-for-international-pages)
+- [Google Blog: How x-default can help you (2023)](https://developers.google.com/search/blog/2023/05/x-default)
+### Language & Region Codes
+Language: ISO 639-1 (2-letter). Region: ISO 3166-1 Alpha 2 (2-letter). Format: `language[-script][-region]`.
+You cannot specify a region code alone. Common mistakes: `en-UK` (should be `en-GB`), `es-419` (not ISO 3166-1). A study found 8.9% of sites using hreflang contain invalid language codes.
+- [Google Search Central: Localized Versions](https://developers.google.com/search/docs/specialty/international/localized-versions)
+- [SE Land: 31% Study](https://searchengineland.com/study-31-of-international-websites-contain-hreflang-errors-395161)
+### Hreflang at Scale (20+ locales)
+With 20 locales, HTML `<head>` hreflang adds ~1.5KB per page for zero user benefit. Sitemap-based hreflang has zero runtime performance impact. `<xhtml:link>` child elements do NOT count toward the 50,000 URL sitemap limit (only `<loc>` elements count).
+John Mueller recommends focusing hreflang on pages receiving wrong-language traffic, not every page: "I wouldn't do it for any of the other pages of the site because it's so complex & hard to manage."
+- [SERoundtable: Child Elements Don't Count](https://www.seroundtable.com/google-child-elements-dont-count-towards-sitemap-url-limit-34377.html)
+- [SERoundtable: Where To Focus Hreflang](https://www.seroundtable.com/using-hreflang-34127.html)
+- [Yoast: hreflang Ultimate Guide](https://yoast.com/hreflang-ultimate-guide/)
+### Google vs Bing
+Bing treats hreflang as a "weak signal." Bing relies on `content-language` meta tag, HTML `lang` attribute, ccTLDs, and server location. Yandex supports hreflang like Google.
+For both engines: implement hreflang (Google/Yandex) + `<html lang="...">` + `<meta http-equiv="content-language">` (Bing).
+- [Digital Ready Marketing: Bing Doesn't Use Hreflang](https://digitalreadymarketing.com/bing-doesnt-use-hreflang-annotation-what-does-it-use/)
+- [Yoast: hreflang Ultimate Guide](https://yoast.com/hreflang-ultimate-guide/)
+---
+## Canonicalization & i18n
+### Self-Referencing Canonicals
+Each locale page must canonical to itself. John Mueller: "Don't use a rel=canonical across languages/countries, only use it on a per-country/language basis."
+Google's docs: "Specify a canonical page in the same language, or the best possible substitute language if a canonical doesn't exist for the same language."
+- [John Mueller: hreflang canonical](https://johnmu.com/hreflang-canonical/)
+- [Google: Consolidate Duplicate URLs](https://developers.google.com/search/docs/crawling-indexing/consolidate-duplicate-urls)
+### Canonical Overrides Hreflang
+Mueller: "If your canonical is pointing somewhere else, Google will follow that and ignore your hreflang annotation." The canonical URL must be one of the URLs in the hreflang set, or all hreflang markup is ignored.
+Google also states: "Google prefers URLs that are part of hreflang clusters for canonicalization" -- when signals align, hreflang strengthens canonical selection.
+- [John Mueller: hreflang canonical](https://johnmu.com/hreflang-canonical/)
+- [SEJ: Hreflang Tags Are Hints](https://www.searchenginejournal.com/google-reminds-that-hreflang-tags-are-hints-not-directives/546428/)
+- [Google: Consolidate Duplicate URLs](https://developers.google.com/search/docs/crawling-indexing/consolidate-duplicate-urls)
+### Near-Duplicate Regional Variants
+Mueller (2023 Office Hours): "If the content is completely the same, and we can't tell any difference, then for simplicity and user experience we may just show one version -- even if hreflang is present."
+Google's duplicate detection runs BEFORE hreflang evaluation. To keep both versions indexed, you need substantive content differences beyond currency symbols.
+- [International Web Mastery: Same-Language Duplicate Pages](https://internationalwebmastery.com/blog/how-google-handles-canonicalization-of-same-language-duplicate-near-duplicate-pages/)
+- [Google: Managing Multi-Regional Sites](https://developers.google.com/search/docs/specialty/international/managing-multi-regional-sites)
+### Pagination Across Locales
+Google: "Don't use the first page of a paginated sequence as the canonical page. Instead, give each page its own canonical URL." Each paginated page in each locale gets self-referencing canonical. `rel="next/prev"` deprecated March 2019.
+- [Google: Pagination Best Practices](https://developers.google.com/search/docs/specialty/ecommerce/pagination-and-incremental-page-loading)
+---
+## International Sitemaps
+### Structure
+Each `<url>` entry includes `<xhtml:link>` alternates for every locale. Requires `xmlns:xhtml="http://www.w3.org/1999/xhtml"` namespace.
+Split sitemaps by content type, not by locale. Splitting by locale creates maintenance problems because every locale sitemap must reference every other locale (reciprocal requirement).
+- [Google Search Central: Localized Versions](https://developers.google.com/search/docs/specialty/international/localized-versions)
+- [Lumar: How Google Handles Hreflang](https://www.lumar.io/office-hours/hreflang/)
+### Size Limits
+50,000 URLs / 50MB uncompressed per sitemap. Only `<loc>` elements count toward the 50K limit. But with 20 hreflang alternates per entry, the 50MB file size limit becomes the bottleneck. Plan for 2,000-5,000 URLs per sitemap when using full hreflang.
+- [Google: Build and Submit a Sitemap](https://developers.google.com/search/docs/crawling-indexing/sitemaps/build-sitemap)
+- [SERoundtable: Sitemap 50,000 Limit](https://www.seroundtable.com/google-sitemap-50-000-limit-based-on-location-urls-not-alternative-urls-33843.html)
+### Submission
+Submit the sitemap index in Search Console AND reference it in robots.txt. Individual child sitemaps can be submitted separately for per-sitemap reporting.
+- [Google: Build and Submit a Sitemap](https://developers.google.com/search/docs/crawling-indexing/sitemaps/build-sitemap)
+### Next.js Caveat
+Next.js `alternates.languages` does NOT automatically include a self-referencing `<xhtml:link>` for the `<loc>` URL. You must explicitly include the `<loc>` URL's own language in the `languages` object.
+- [Next.js Docs: sitemap.xml](https://nextjs.org/docs/app/api-reference/file-conventions/metadata/sitemap)
+---
+## URL Structure
+### Strategies Compared
+Google treats subdirectories and subdomains equivalently. Mueller: "From our point of view...they say subdomains and subdirectories are essentially equivalent."
+URL parameters (`?lang=en`) are explicitly "Not recommended" per Google docs.
+- [Google: Managing Multi-Regional Sites](https://developers.google.com/search/docs/specialty/international/managing-multi-regional-sites)
+### Default Language
+Mueller recommends: set `/` as x-default, put each language in its own prefix. Without marking `/` as x-default, "to Google it can look like '/' is a separate page from the others."
+- [Google Blog: x-default](https://developers.google.com/search/blog/2023/05/x-default)
+- [Google Blog: Creating the Right Homepage](https://developers.google.com/search/blog/2014/05/creating-right-homepage-for-your)
+### Content Negotiation / IP Redirects
+Google strongly advises against locale-adaptive pages. Googlebot crawls from US IPs and does not send Accept-Language headers. Separate URLs + hreflang are required.
+- [Google: Locale-Adaptive Pages](https://developers.google.com/search/docs/specialty/international/locale-adaptive-pages)
+### Trailing Slash Consistency
+Mueller: trailing slash is "a significant part of the URL and will change the URL if it's there or not." Pick one format for all locale paths, internal links, canonicals, hreflang, and sitemaps.
+Mueller (2025): "Consistency is the biggest technical SEO factor."
+- [SERoundtable: Consistency Is The Biggest Technical SEO Factor](https://www.seroundtable.com/google-consistency-seo-40427.html)
+### Search Console Geotargeting
+The International Targeting report is deprecated. Google now relies entirely on hreflang, content language analysis, and linking patterns. You can add subdirectory properties for per-locale reporting.
+- [Google Support: International Targeting Deprecated](https://support.google.com/webmasters/answer/12474899?hl=en)
+### Framework Locale Modes
+Use `localePrefix: 'always'` (next-intl) or equivalent. Never hide locale from URLs -- Google needs unique URLs per language. Using `'never'` mode disables alternate links entirely.
+- [next-intl: Routing Configuration](https://next-intl.dev/docs/routing/configuration)
+- [Next.js Discussion #18419](https://github.com/vercel/next.js/discussions/18419)
+---
+## Content Quality Across Locales
+### Auto-Translated Content (2025 Stance)
+Google removed longstanding guidance advising against auto-translated content in mid-2025. Current stance: "Our policies do not strictly define content that has been translated by AI as spam." The scaled content abuse policy mentions translation as a possible vector, but does not ban it.
+Reddit scaled AI translations to 35+ languages with Google's knowledge. The key distinction is intent and quality, not the method.
+- [Google Spam Policies](https://developers.google.com/search/docs/essentials/spam-policies)
+- [Glenn Gabe: Auto-Translating Content](https://www.gsqi.com/marketing-blog/auto-translating-content-google-scaled-content-abuse/)
+- [SE Land: Reddit AI Translations](https://searchengineland.com/google-comments-on-reddits-use-of-ai-to-translate-its-pages-456908)
+### Thin Locale Pages
+Google: "Localized versions of a page are only considered duplicates if the main content of the page remains untranslated." Pages with only translated boilerplate get clustered as duplicates.
+Do NOT use noindex for unwanted locale pages (wastes crawl budget). Do NOT canonical cross-locale (conflicts with hreflang). Best approach: don't create locale pages you can't make genuinely helpful.
+- [Google: Localized Versions](https://developers.google.com/search/docs/specialty/international/localized-versions)
+- [Google: Crawl Budget Management](https://developers.google.com/search/docs/crawling-indexing/large-site-managing-crawl-budget)
+### Helpful Content System Impact
+Merged into core ranking March 2024. Site-wide signal: "any content -- not just unhelpful content -- on sites determined to have relatively high amounts of unhelpful content overall is less likely to perform well in Search."
+Low-quality translated pages can drag down the entire site. This is the strongest argument against creating locale pages that aren't genuinely helpful.
+- [Google Blog: Helpful Content Update](https://developers.google.com/search/blog/2022/08/helpful-content-update)
+- [Amsive: What Changed in 2024](https://www.amsive.com/insights/seo/googles-helpful-content-update-ranking-system-what-happened-and-what-changed-in-2024/)
+### Partial Translation
+Google: "Translating only the boilerplate text of your pages while keeping the bulk of your content in a single language...can create a bad user experience." Google uses visible content (not lang attribute) to determine page language.
+Translate ALL content on a page if you create a locale version. Untranslated metadata (title, description) in the wrong language reduces CTR.
+- [Google: Managing Multi-Regional Sites](https://developers.google.com/search/docs/specialty/international/managing-multi-regional-sites)
+### Crawl Budget
+Only a concern for 1M+ pages or 10K+ pages changing daily. But alternate URLs (hreflang targets) do consume crawl budget. Broken hreflang links waste budget AND invalidate signals.
+- [Google: Crawl Budget Management](https://developers.google.com/search/docs/crawling-indexing/large-site-managing-crawl-budget)
+- [Google Blog: Crawl Budget](https://developers.google.com/search/blog/2017/01/what-crawl-budget-means-for-googlebot)
+### Locale-Specific Signals
+Google identifies audience via: "local addresses and phone numbers on the pages, the use of local language and currency, links from other local sites, or signals from your Business Profile."
+- [Google: Managing Multi-Regional Sites](https://developers.google.com/search/docs/specialty/international/managing-multi-regional-sites)

package/{dist/src/orgrt/role-skills/seo-specialist.md → org-skills/seo-specialist/SKILL.md} RENAMED Viewed

@@ -1,3 +1,11 @@
+---
+name: seo-specialist
+description: "SEO Specialist role best practices. Improves organic visibility through technical SEO, content optimization, and authority building — increasingly across both traditional search and AI-driven answer engines."
+tags: ["writing","marketing"]
+tools: []
+license: Apache-2.0
+source: https://github.com/monoes/monomind
+---
 # SEO Specialist — Best Practices
 ## Focus

package/org-skills/shap-model-explainability/LICENSE.txt ADDED Viewed

@@ -0,0 +1,21 @@
+MIT License
+Copyright (c) 2025 K-Dense Inc.
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/org-skills/shap-model-explainability/SKILL.md ADDED Viewed

@@ -0,0 +1,298 @@
+---
+name: shap-model-explainability
+description: "Explain and audit machine-learning predictions with SHAP. Use for selecting SHAP explainers and maskers, computing and validating feature attributions, handling multi-output explanations, and producing local or global SHAP visualizations."
+tags: ["ai-ml","data"]
+tools: []
+license: MIT
+source: https://github.com/K-Dense-AI/scientific-agent-skills
+source_path: "skills/shap"
+source_commit: 49c6e97775eaa18ba791bebe23162a70ae601c18
+---
+# SHAP
+Use SHAP to describe how a fitted predictive model maps inputs to outputs. Work from the modern `shap.Explanation` API, make the explained output and background distribution explicit, and validate every explanation before interpreting it.
+This skill is aligned with **SHAP 0.52.0** (released 2026-05-28). That release requires Python 3.12 or newer.
+## Operating Rules
+1. Explain a fixed, evaluated model; do not use SHAP as a substitute for predictive validation.
+2. Use held-out or clearly labeled analysis rows for explanations. Choose background rows only from an appropriate training or reference population.
+3. State the explained output: regression value, raw margin, probability, log loss, logit, or another model method.
+4. Keep explanations as `shap.Explanation` objects. Call `explainer(X)`; use `.shap_values(X)` only when maintaining legacy code.
+5. For multi-output models, select one output before using tabular plots: `explanation[..., output_index]`.
+6. Check `base_values + values.sum(...)` against the exact model output being explained.
+7. Treat SHAP as a description of model behavior under a masking/background choice. It does not establish causality, fairness, recourse, or scientific mechanism.
+8. Never silence an additivity failure until input shape, preprocessing, model version, output space, and row ordering have been checked.
+9. Do not load untrusted pickle, joblib, model, or explainer artifacts; those formats can execute code during deserialization.
+## Install
+Create an isolated environment and pin the documented release:
+```bash
+uv venv --python 3.12
+source .venv/bin/activate
+uv pip install "shap[plots]==0.52.0"
+```
+`shap[plots]` installs the plotting dependencies. Add the fitted model's package at a version compatible with the project. For older Python compatibility, read [references/migration.md](references/migration.md) instead of silently installing a different SHAP release.
+Confirm the environment before debugging an API mismatch:
+```python
+import platform
+import shap
+print("Python:", platform.python_version())
+print("SHAP:", shap.__version__)
+```
+## Standard Workflow
+### 1. Define the explanation target
+Record:
+- model and preprocessing version;
+- exact callable or model method being explained;
+- output name/index and units;
+- evaluation rows;
+- background/reference population;
+- masker and explainer algorithm;
+- SHAP and model-library versions.
+For classifiers, decide whether the task needs raw margins or probabilities. Defaults differ by model family; never infer units from the plot color or sign.
+### 2. Select an explainer and masker
+Start with `shap.Explainer(model, masker)` when automatic dispatch is sufficient. Instantiate a specialized explainer when its assumptions or output controls matter.
+| Situation | Preferred choice | Important constraint |
+|---|---|---|
+| Supported tree ensemble | `TreeExplainer` | `model_output="probability"` and `"log_loss"` require interventional masking and background data |
+| Linear model | `LinearExplainer` | The masker determines interventional versus correlation-aware behavior |
+| Small feature space | `ExactExplainer` | Cost grows quickly with unconstrained feature count |
+| General tabular callable | `PermutationExplainer` | Budget at least one full forward/reverse permutation |
+| Hierarchical feature groups, text, or image | `PartitionExplainer` | The partition tree changes the cooperative game |
+| Differentiable neural network | `DeepExplainer` or `GradientExplainer` | Framework support, output shape, and background choice require testing |
+| Legacy Kernel SHAP workflow | `KernelExplainer` | Usually much slower than model-specific methods |
+Use the detailed decision guide in [references/explainers.md](references/explainers.md). Use [references/data-maskers.md](references/data-maskers.md) when features are correlated, structured, sparse, or semantically grouped.
+### 3. Compute a modern `Explanation`
+This complete binary-classification example uses an explicit background and selects the positive-class output:
+```python
+import numpy as np
+import shap
+from sklearn.datasets import load_breast_cancer
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.model_selection import train_test_split
+X, y = load_breast_cancer(as_frame=True, return_X_y=True)
+X_train, X_test, y_train, y_test = train_test_split(
+    X,
+    y,
+    test_size=0.2,
+    stratify=y,
+    random_state=7,
+)
+model = RandomForestClassifier(
+    n_estimators=200,
+    min_samples_leaf=3,
+    random_state=7,
+    n_jobs=-1,
+).fit(X_train, y_train)
+background = shap.sample(X_train, 100, random_state=7)
+explainer = shap.Explainer(model, background, algorithm="tree")
+all_outputs = explainer(X_test)
+# sklearn tree classifiers expose one output per class.
+positive = all_outputs[..., 1]
+assert positive.values.shape == X_test.shape
+reconstructed = np.asarray(positive.base_values) + positive.values.sum(axis=1)
+expected = model.predict_proba(X_test)[:, 1]
+np.testing.assert_allclose(reconstructed, expected, rtol=1e-5, atol=1e-6)
+shap.plots.beeswarm(positive, max_display=15)
+shap.plots.waterfall(positive[0], max_display=15)
+```
+Output shape is model-dependent:
+- one tabular output: `(samples, features)`;
+- multiple tabular outputs: `(samples, features, outputs)`;
+- multiple model inputs: often a list of arrays or explanations;
+- image/text explanations: feature axes follow the input representation, with output selection on the final axis when present.
+Do not use the pre-0.45 pattern `values[class_index]` for a modern multi-output array. Use `values[..., class_index]` or slice the `Explanation` itself.
+### 4. Control tree output semantics when needed
+For a supported tree classifier, probability-space explanations must be explicit:
+```python
+background = shap.sample(X_train, 200, random_state=7)
+explainer = shap.TreeExplainer(
+    model,
+    data=background,
+    feature_perturbation="interventional",
+    model_output="probability",
+)
+probability_exp = explainer(X_test)
+```
+In SHAP 0.52:
+- `feature_perturbation="auto"` uses interventional semantics when background data is supplied and tree-path-dependent semantics otherwise;
+- probability and log-loss output modes are supported only with interventional semantics;
+- pass `approximate=True` to `explainer(X, approximate=True)` if deliberately using the lower-fidelity tree approximation; do not pass it to the constructor.
+### 5. Use a model-agnostic callable deliberately
+Pass the exact callable whose outputs will be interpreted:
+```python
+masker = shap.maskers.Independent(background, max_samples=100)
+explainer = shap.Explainer(
+    model.predict_proba,
+    masker,
+    algorithm="permutation",
+    output_names=[str(label) for label in model.classes_],
+    seed=7,
+)
+budget = 2 * X_test.shape[1] + 1
+all_outputs = explainer(X_test.iloc[:20], max_evals=budget)
+positive = all_outputs[..., 1]
+```
+Increase `max_evals` to average over more permutations when estimates are unstable. Keep the seed, background sample, and evaluation budget in the report.
+### 6. Visualize the question, not merely the available plot
+| Question | Plot |
+|---|---|
+| Which features have the largest average attribution magnitude? | `shap.plots.bar(exp)` |
+| How do direction, magnitude, and observed values vary globally? | `shap.plots.beeswarm(exp)` |
+| Why did one prediction differ from its baseline? | `shap.plots.waterfall(exp[i])` |
+| How does one feature's attribution vary over its values? | `shap.plots.scatter(exp[:, feature])` |
+| Do explanations form sample-level patterns? | `shap.plots.heatmap(exp)` |
+| How do predefined cohorts differ descriptively? | `shap.plots.bar(exp.cohorts(labels).abs.mean(0))` |
+| Which tokens or image regions contribute to an output? | `shap.plots.text(exp)` or `shap.plots.image(exp)` |
+Read [references/plots.md](references/plots.md) before customizing or saving figures.
+### 7. Report limitations with results
+At minimum, report:
+- output and units;
+- baseline/reference population;
+- explainer and masker;
+- sample count and selection;
+- output index/name;
+- additivity error or applicable approximation diagnostics;
+- known correlated/grouped features;
+- whether results are local, aggregated, or cohort-specific;
+- a clear non-causal statement.
+## Common Tasks
+### Global and local analysis
+Use global plots to locate important patterns, scatter plots to inspect those patterns, and local plots to investigate selected rows. Do not select only visually dramatic rows without documenting the selection rule.
+### Multiclass models
+Set `output_names` where possible, inspect `explanation.output_names`, and slice an output before plotting:
+```python
+class_exp = explanation[..., "class_name"]
+# or
+class_exp = explanation[..., class_index]
+```
+Never average signed attributions across classes. For cross-class comparison, preserve the same model, rows, background, output space, and aggregation.
+### Cohorts, subgroup analysis, and fairness
+SHAP can compare how a model uses features across cohorts, but this is not a fairness test. A protected feature with small SHAP magnitude does not rule out proxy discrimination, and removing a protected feature does not establish fairness. Pair attribution analysis with performance, calibration, error-rate, and domain-appropriate fairness metrics.
+See [references/workflows.md](references/workflows.md) for cohort construction, model comparison, error analysis, log-loss explanations, monitoring, and production records.
+### Text and images
+Use domain maskers rather than treating tokens or pixels as ordinary independent columns:
+- `shap.maskers.Text(tokenizer)` with `PartitionExplainer` for token groups;
+- `shap.maskers.Image(...)` with `PartitionExplainer` for image regions;
+- restrict expensive multi-output models with `outputs=...`.
+Read [references/modalities.md](references/modalities.md) for current examples and output-shape guidance.
+## Troubleshooting Order
+1. Print Python, SHAP, model-library, NumPy, and framework versions.
+2. Verify the model receives exactly the same transformed columns, order, dtype, and missing-value representation used during fitting.
+3. Print `values.shape`, `base_values.shape`, `data.shape`, `feature_names`, and `output_names`.
+4. Confirm the selected output and output units.
+5. Recompute predictions on the same rows in the same order.
+6. Test a smaller batch and representative background.
+7. Only then investigate package-specific compatibility or approximation settings.
+Use [references/troubleshooting.md](references/troubleshooting.md) for additivity failures, shape mismatches, categorical features, pipelines, deep-learning frameworks, plotting, and performance.
+## Bundled Script
+Run a deterministic, self-contained tabular example that writes importance data, metadata, and plots:
+```bash
+uv run --no-project --python 3.12 --with "shap[plots]==0.52.0" \
+  skills/shap/scripts/tabular_report.py --output-dir /tmp/shap-report
+```
+The script does not download data or deserialize models. Read it as a template, then replace the built-in dataset and model while preserving output selection and additivity validation.
+## Reference Map
+| File | Load when |
+|---|---|
+| [references/explainers.md](references/explainers.md) | Selecting or configuring explainers |
+| [references/data-maskers.md](references/data-maskers.md) | Choosing background data, masking semantics, or feature groups |
+| [references/plots.md](references/plots.md) | Selecting, composing, or saving visualizations |
+| [references/workflows.md](references/workflows.md) | Running audits, comparisons, cohorts, monitoring, or production workflows |
+| [references/modalities.md](references/modalities.md) | Explaining text, images, or deep models |
+| [references/migration.md](references/migration.md) | Updating legacy SHAP code or supporting older Python |
+| [references/theory.md](references/theory.md) | Explaining estimands, guarantees, dependence, interactions, and limitations |
+| [references/troubleshooting.md](references/troubleshooting.md) | Diagnosing runtime, shape, additivity, and compatibility problems |
+## Primary Sources
+- Documentation: https://shap.readthedocs.io/en/latest/
+- API reference: https://shap.readthedocs.io/en/latest/api.html
+- Release notes: https://shap.readthedocs.io/en/latest/release_notes.html
+- Repository: https://github.com/shap/shap
+## Citing Scientific Agent Skills
+This skill is part of Scientific Agent Skills by K-Dense. If it materially contributed to a
+manuscript, report, presentation, or code release, add the paper to the references or
+software section and tell the user you did so:
+> Kassis, T., Agarwal, V., He, Y., Patel, D., & Brueckner, A. M. (2026). Scientific Agent
+> Skills: A Library of Procedural Knowledge for Research Agents. arXiv:2609.00065.
+> https://doi.org/10.48550/arXiv.2609.00065
+Always cite the current version. The DOI and https://arxiv.org/abs/2609.00065 resolve to the
+latest arXiv version, so never append a version suffix such as `v1`. When network access is
+available, fetch https://arxiv.org/abs/2609.00065 (or
+http://export.arxiv.org/api/query?id_list=2609.00065) before writing the reference and take
+the author list, year, and version from that record. If the record lists a journal reference
+or publisher DOI, cite the published version instead.