[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7EBD9VM772XEYWRTB5THZPC":3,"skill-same:01M7EBD9VM772XEYWRTB5THZPC":62},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":30,"commitSha":31,"skillDirectory":32,"forksCount":14,"repositoryHtmlUrl":33,"repositoryHomepageUrl":33,"ownerHtmlUrl":34,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":35,"install":36,"resources":49},"01M7EBD9VM772XEYWRTB5THZPC","bio-admet-prediction",null,"Predicts ADMET properties using ADMETlab 3.0 (119 platform features, including 77 prediction models with modeled-endpoint uncertainty), ADMET-AI, DeepChem MolNet, and chemprop D-MPNN with explicit handling of OECD QSAR principles, applicability domain assessment, calibration, hERG\u002FCYP\u002FAMES endpoints, and PAINS \u002F Lipinski \u002F Ro5 \u002F Veber \u002F BBB druglikeness filters. Use when filtering compounds for drug-likeness, prioritizing leads by predicted safety, or building an in-house ADMET QSAR model.","Mr-Nobody-Anonymous\u002FAll-skills","Mr-Nobody-Anonymous","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F184253119?v=4","01M7AEXFJHKTYQW5R0ZCQQTWFV","anonymous","01M7C00M3GQR4QKP9GZA42GFAK",0,"MIT","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-admet-prediction\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: bio-admet-prediction\ndescription: Predicts ADMET properties using ADMETlab 3.0 (119 platform features, including 77 prediction models with modeled-endpoint uncertainty), ADMET-AI, DeepChem MolNet, and chemprop D-MPNN with explicit handling of OECD QSAR principles, applicability domain assessment, calibration, hERG\u002FCYP\u002FAMES endpoints, and PAINS \u002F Lipinski \u002F Ro5 \u002F Veber \u002F BBB druglikeness filters. Use when filtering compounds for drug-likeness, prioritizing leads by predicted safety, or building an in-house ADMET QSAR model.\ntool_type: python\nprimary_tool: ADMETlab\nsource: \"https:\u002F\u002Fgithub.com\u002FGPTomics\u002FbioSkills\"\nsource_repository: \"GPTomics\u002FbioSkills\"\nsource_path: \"chemoinformatics\u002Fadmet-prediction\u002FSKILL.md\"\nlicense: \"MIT\"\nimported_at: \"2026-09-20\"\n---\n\n## Version Compatibility\n\nReference examples tested with: RDKit 2024.09+, requests 2.31+, DeepChem 2.8+, chemprop 2.0+ (note major API change from 1.x), admet-ai 1.3+, pandas 2.2+.\n\nBefore using code patterns, verify installed versions match. If versions differ:\n- Python: `pip show \u003Cpackage>` then `help(module.function)` to check signatures\n\nIf code throws ImportError, AttributeError, or TypeError, introspect the installed\npackage and adapt the example to match the actual API rather than retrying.\n\n# ADMET Prediction\n\nPredict absorption, distribution, metabolism, excretion, and toxicity properties of drug candidates. ADMET prediction underpins lead selection and de-risking; calibrated, applicability-domain-aware predictions distinguish a working filter from a costly false-confidence rejection. Modern best practice combines online services (ADMETlab 3.0 with uncertainty estimates), open-source models (chemprop D-MPNN), and rule-based filters (Lipinski \u002F Veber \u002F BBB heuristics) -- each with known failure modes.\n\nFor PAINS \u002F Brenk \u002F structural alerts, see `chemoinformatics\u002Fsubstructure-search`. For QSAR model building from in-house data, see `chemoinformatics\u002Fqsar-modeling`.\n\n## ADMET Model Taxonomy\n\n| Tool | Endpoints | Architecture | Uncertainty | Access | Fails when |\n|------|-----------|--------------|-------------|--------|------------|\n| ADMETlab 3.0 | 119 reported features: 77 prediction models, 34 computed properties, 8 rules | Multi-task DMPNN + descriptors for modeled endpoints | Evidential uncertainty for modeled endpoints | Web service; hosted API documented by the authors | Outside training distribution; metals; macrocycles |\n| ADMET-AI | TDC-derived ADMET tasks; inspect installed model metadata | Chemprop D-MPNN | Inspect version-specific outputs; do not assume calibrated uncertainty | Python package | v2 package predictions differ from the v1 paper\u002Fserver |\n| DeepChem MolNet | Dataset-dependent tasks including Tox21, ToxCast, and ClinTox | Model-dependent | Model-dependent | Python package | Coverage and uncertainty depend on the selected dataset\u002Fmodel |\n| pkCSM | Service-defined ADMET endpoints | Graph signatures + ML | Inspect current service output | Web service | Applicability domain and service contract must be checked |\n| SwissADME | Physchem, pharmacokinetics, drug-likeness, and medchem outputs | Published models and rules | None advertised | Web service (no public API) | Automated access is restricted by its terms |\n| ProTox-3.0 | 61 toxicity models\u002Fendpoints | RF\u002FDNN + fingerprints, similarity, and pharmacophore methods | Confidence score | Web service \u002F sample API | Toxicity only; reports LD50 and toxicity class |\n| ADMETpredictor (Simulations Plus) | ~140 | Proprietary | Per-prediction | Commercial | License cost |\n| FAF-Drugs4 | filters | Rule-based | None | Web | Static rules |\n| chemprop (in-house) | User-defined | D-MPNN ± descriptors | Ensemble and other estimators; optional calibration | Python package | Requires suitable training and calibration data |\n\n**Decision:** For batch screening with no in-house data, **ADMETlab 3.0** provides 119 reported platform features and uncertainty for modeled endpoints through its web service and hosted API; verify the live API documentation before automating access. For a sufficiently large, relevant in-house endpoint dataset, benchmark a **chemprop D-MPNN**, descriptors, and simpler baselines under a deployment-relevant split rather than assuming a universal sample-size threshold. Shan et al. (2022) reported an AUC of 0.956 for a D-MPNN combined with 206 MOE descriptors on their random-split hERG benchmark.\n\n## Decision Tree by Scenario\n\n| Scenario | Workflow | Reasoning |\n|----------|----------|-----------|\n| Library triage, no in-house data | ADMETlab 3.0 API batch | Broad platform coverage plus modeled-endpoint uncertainty |\n| Single endpoint, adequate in-house data | Benchmark chemprop D-MPNN, descriptors, and simpler baselines | Select by prospective or deployment-relevant validation |\n| Need calibrated probabilities | chemprop with ensemble + Platt | Native deep learning rarely calibrated |\n| FDA \u002F regulatory submission | OECD-compliant QSAR with AD | See OECD principles below |\n| Quick annotation for VS | Lipinski, Veber, and QED reported separately | Rank or annotate; do not impose a universal QED gate |\n| BBB penetration | Simple screen: TPSA \u003C= 90, MW \u003C= 500, HBD \u003C= 3 | Repository heuristic; not the six-factor CNS MPO |\n| Cardiotox liability | ADMETlab hERG + ProTox-3.0 cardiotoxicity + literature check | Both hosted endpoints model hERG blockade; compare applicability domains and assay definitions |\n| Drug-drug interaction (CYP) | CYP1A2\u002F2C9\u002F2C19\u002F2D6\u002F3A4 inhibitor + substrate | Standard set of 5 CYPs |\n\n## OECD QSAR Principles (5 Pillars)\n\nFor regulatory-grade ADMET QSAR (REACH, ECHA, FDA submissions), models must satisfy:\n\n1. **Defined endpoint** -- specific bioassay, units, conditions\n2. **Unambiguous algorithm** -- reproducible model + code\n3. **Defined applicability domain (AD)** -- where the model is valid\n4. **Appropriate statistical validation** -- external test set, cross-validation\n5. **Mechanistic interpretation** -- biological \u002F chemical rationale\n\nFor non-regulatory work, AD assessment is still critical. The OECD's *applicability domain* is the workhorse: predictions outside the AD are unreliable, but operational AD measures (leverage, kNN, conformal prediction) often disagree.\n\n## Applicability Domain Methods\n\n| Method | Definition | Flags out-of-AD when |\n|--------|-----------|----------------------|\n| kNN distance | Mean distance to k nearest neighbors in training set | > training-set distribution P95 |\n| Leverage (Williams) | Hat-matrix diagonal | > 3p\u002Fn (p = features, n = compounds) |\n| Density (KDE on PCA) | Density in feature space | \u003C density of training set P5 |\n| Conformal prediction | Per-prediction confidence interval | Interval > tolerance |\n| Bayesian variance | Ensemble or MC-dropout variance | > training-set variance P95 |\n\nFor deep-learning ADMET, **conformal prediction** can provide calibrated prediction sets or intervals when its exchangeability and calibration assumptions are appropriate (McShane et al. 2024).\n\n## ADMETlab 3.0 API\n\nADMETlab 3.0 reports 119 platform features: 77 prediction models, 34 computed physicochemical properties, and 8 medicinal-chemistry rules. The modeled endpoints include prediction uncertainty; do not imply that computed properties and rules have model uncertainty.\n\n**Goal:** Obtain the platform's 119 features for a batch of SMILES, including uncertainty for the 77 modeled endpoints, using the hosted service.\n\n**Approach:** Follow the live ADMETlab 3.0 API tutorial to wash molecules, submit batch predictions, and retrieve the returned results. The 2024 paper documents API\u002Fbatch support and modeled-endpoint uncertainty. Obtain current rate limits, routes, payloads, task identifiers, and output contracts from the live official documentation rather than attributing them to the paper or hard-coding an unofficial example.\n\n```python\nimport pandas as pd\n\n# After submitting with the current official API example, load its CSV output.\nresults = pd.read_csv('admetlab3_results.csv')\n# Preserve the uncertainty columns and task identifier in downstream reports.\n```\n\nADMETlab 3.0 endpoints (sample):\n- Absorption: Caco-2 permeability (logPapp), HIA (%), Pgp inhibitor\u002Fsubstrate, MDCK\n- Distribution: BBB+, PPB (%), VDss (L\u002Fkg), Fu (fraction unbound)\n- Metabolism: CYP1A2\u002F2C9\u002F2C19\u002F2D6\u002F3A4 inhibitor \u002F substrate\n- Excretion: CL (mL\u002Fmin\u002Fkg), T1\u002F2 (h)\n- Toxicity: hERG, AMES, hepatotoxicity (DILI), carcinogenicity, immunotoxicity, mutagenicity, respiratory, skin, eye, cardiotoxicity, mitochondrial, NR-AR, NR-ER, SR-MMP\n- Drug-likeness: Lipinski, Veber, Ghose, Egan, Muegge, QED, SAscore\n\n## chemprop D-MPNN for Custom Endpoints\n\nWhen in-house data is available, train a target-specific model. chemprop provides a widely used open-source D-MPNN architecture with atom\u002Fbond features and optional molecular descriptors; benchmark it against appropriate baselines on the project's data.\n\n**Goal:** Train a target-specific ADMET classifier or regressor on in-house bioassay data.\n\n**Approach:** Use the installed Chemprop 2.x CLI with a scaffold split, a release-supported descriptor featurizer, replicated models, and an explicit prediction-time uncertainty\u002Fcalibration workflow.\n\n```python\n# Chemprop 2.2 CLI; verify flags against the installed release.\n# chemprop train --data-path data.csv --task-type classification \\\n#                --save-dir model_dir --split-type scaffold_balanced \\\n#                --molecule-featurizers v1_rdkit_2d_normalized \\\n#                --num-replicates 5 --ensemble-size 5\n# chemprop predict --test-path test.csv --model-paths model_dir \\\n#                  --uncertainty-method ensemble --preds-path predictions.csv\n\n# Or chemprop 2.x programmatic API (full programmatic API documented at chemprop.readthedocs.io)\n# See chemoinformatics\u002Fqsar-modeling for the full chemprop 2.x training pipeline.\n```\n\n**Key:** Replicates and an ensemble estimator produce an uncertainty estimate, not automatic calibration. Fit and evaluate a documented calibrator on a separate calibration set when calibrated probabilities or intervals are required. Descriptor benefit must be demonstrated on the intended endpoint.\n\n## hERG Cardiotoxicity Endpoint\n\nhERG (KCNH2) blockade can contribute to QT prolongation and Torsades de Pointes and is an important non-clinical cardiac-safety endpoint. Follow the current ICH S7B\u002FE14 and regulator-specific guidance applicable to the program rather than treating one model output as a regulatory conclusion.\n\n| Model | Architecture | Training data | AUC | Reference |\n|-------|--------------|---------------|-----|-----------|\n| Shan et al. D-MPNN + MOE | D-MPNN + 206 MOE descriptors | 7,889 compounds | 0.956 (random split) | Shan 2022 |\n| CardioTox-net | Five DL base representations + neural meta-ensemble | BindingDB, ChEMBL, and literature | 0.930 (10-fold meta-validation) | Karim 2021 |\n| ADMETlab 3.0 hERG | DMPNN multi-task | Internal | 0.92 (reported) | Fu 2024 |\n| ProTox-3.0 cardiotoxicity | RF-based classifier | 5,252 ChEMBL compounds with hERG IC50\u002FKi | 0.86 CV; 0.95 external | Banerjee 2024 |\n\n**Interpretation:** A single-model probability > 0.5 is NOT a kill signal. Triangulate multiple hERG-specific models and a literature search. ProTox-3.0 calls the endpoint cardiotoxicity, but its model specifically predicts small-molecule hERG blockers; it should not be treated as an independent non-hERG mechanism. Consider exposure relative to measured hERG potency and confirm important decisions experimentally rather than applying a universal safe\u002Funsafe IC50 cutoff.\n\n## CYP Inhibition (DDI Risk)\n\n5 CYP isoforms cover most clinically relevant DDIs:\n\n| CYP | Substrates (drugs) | Inhibitor flag if predicted prob | Action |\n|-----|--------------------|-----------------------------------|--------|\n| CYP3A4 | many drug classes | Model-specific threshold | Interpret inhibitor and substrate assays separately |\n| CYP2D6 | beta-blockers, antidepressants | Model-specific threshold | Include polymorphism and exposure context |\n| CYP2C9 | warfarin, NSAIDs | Model-specific threshold | Evaluate clinical substrate\u002Fexposure context |\n| CYP2C19 | PPIs, clopidogrel | Model-specific threshold | Include polymorphism and assay context |\n| CYP1A2 | caffeine, theophylline | Model-specific threshold | Include induction, diet, and smoking context |\n\n## PAINS, BRENK, REOS Filters\n\nADMET prediction is separate from structural alerts; combine. See `chemoinformatics\u002Fsubstructure-search` for PAINS\u002FBRENK\u002FREOS pattern catalogs.\n\n```python\nfrom rdkit.Chem.FilterCatalog import FilterCatalog, FilterCatalogParams\n\ndef alerts(mol, catalogs=('PAINS_A', 'BRENK', 'ZINC')):\n    params = FilterCatalogParams()\n    for cat in catalogs:\n        params.AddCatalog(getattr(FilterCatalogParams.FilterCatalogs, cat))\n    catalog = FilterCatalog(params)\n    hits = catalog.GetMatches(mol)\n    return [h.GetDescription() for h in hits]\n```\n\n## Lipinski \u002F Veber \u002F Drug-Likeness\n\nSee `chemoinformatics\u002Fmolecular-descriptors` for full physchem table. Quick filter:\n\n```python\nfrom rdkit.Chem import Descriptors, Lipinski, QED\n\ndef druglike_score(mol):\n    mw = Descriptors.MolWt(mol)\n    logp = Descriptors.MolLogP(mol)\n    hbd = Lipinski.NumHDonors(mol)\n    hba = Lipinski.NumHAcceptors(mol)\n    tpsa = Descriptors.TPSA(mol)\n    rotbonds = Lipinski.NumRotatableBonds(mol)\n    qed = QED.qed(mol)\n\n    lipinski_violations = sum([mw > 500, logp > 5, hbd > 5, hba > 10])\n    veber_pass = rotbonds \u003C= 10 and tpsa \u003C= 140\n    bbb_simple_screen = tpsa \u003C= 90 and mw \u003C= 500 and hbd \u003C= 3\n\n    return {'MW': mw, 'LogP': logp, 'HBD': hbd, 'HBA': hba,\n            'TPSA': tpsa, 'RotBonds': rotbonds, 'QED': qed,\n            'Lipinski_violations': lipinski_violations,\n            'Veber_pass': veber_pass, 'BBB_simple_screen': bbb_simple_screen}\n```\n\n## Per-Tool Failure Modes\n\n### ADMETlab 3.0 -- out-of-distribution prediction\n\n**Trigger:** Chemistry materially unlike the service's documented training\u002Fapplicability domain, such as metal-containing complexes, many peptides, PROTACs, or unusual macrocycles.\n\n**Mechanism:** ADMETlab training set is drug-like organic molecules. Predictions on PROTACs, macrocycles, peptides extrapolate.\n\n**Symptom:** High reported uncertainty, disagreement with neighbors or orthogonal models, or unstable conclusions under reasonable preprocessing.\n\n**Fix:** Check uncertainty band; if interval is broad, do not trust point estimate. For PROTACs \u002F macrocycles, prefer literature-derived experimental data.\n\n### hERG D-MPNN -- training data bias\n\n**Trigger:** Compound is novel chemotype not in training set (drug-like but in unexplored region).\n\n**Mechanism:** D-MPNN learns local chemical features; for genuinely new scaffolds, extrapolation is unreliable.\n\n**Symptom:** Model predicts hERG- (false negative) for compound that experimentally inhibits.\n\n**Fix:** Use ensemble + applicability-domain assessment (kNN distance, ensemble variance). If kNN distance to training set > P95, treat prediction as low-confidence.\n\n### CYP3A4 inhibitor + substrate ambiguity\n\n**Trigger:** Model trained on either inhibitor OR substrate; predictions confused.\n\n**Mechanism:** CYP3A4 inhibitors and substrates have similar SAR; many compounds are both.\n\n**Symptom:** Both classes report > 0.5.\n\n**Fix:** Two separate models (inhibitor model, substrate model); compounds that score high in both are flagged for in vitro confirmation.\n\n### SwissADME -- no API\n\n**Trigger:** Wanting to batch programmatically.\n\n**Mechanism:** SwissADME's terms restrict automated crawler\u002Fdata-retrieval access, and no public API is documented.\n\n**Symptom:** No programmatic access; manual web upload only.\n\n**Fix:** Use a currently documented programmatic service and follow its access policy; for ADMETlab 3.0, verify the live API tutorial before writing a client.\n\n### PAINS as a kill filter\n\n**Trigger:** Treating PAINS_A match as a categorical exclusion.\n\n**Mechanism:** PAINS is calibrated against HTS assay-interference; matches do NOT predict failed drug development.\n\n**Symptom:** Library purged of valid leads (curcumin analogs, polyphenol natural products).\n\n**Fix:** Flag PAINS for orthogonal-assay confirmation; do not exclude pre-emptively. See substructure-search for details.\n\n### Class-imbalanced AMES dataset\n\n**Trigger:** Training\u002Fpredicting AMES mutagenicity.\n\n**Mechanism:** Public AMES datasets can be imbalanced and differ in assay definition and curation; aggregate accuracy can therefore be misleading.\n\n**Symptom:** Model reports high accuracy but predicts negative for all.\n\n**Fix:** Report class balance and use suitable metrics such as PR-AUC, ROC-AUC, MCC, or balanced accuracy. Compare class weighting or resampling inside training folds without leaking validation\u002Ftest data.\n\n## Reconciliation Across Models\n\nWhen ADMETlab, ProTox-3.0, and an independently trained chemprop model disagree on hERG:\n- All predict hERG+ -> higher concern; plan in vitro patch-clamp\n- Results disagree -> inspect applicability domains, activity thresholds, and assay definitions before deciding\n- All predict hERG- -> lower concern, but still consider in vitro screening for clinical candidates and novel chemotypes\n- Do not count correlated models as independent evidence merely because they are hosted by different services\n\n## Common Errors\n\n| Symptom | Cause | Fix |\n|---------|-------|-----|\n| ADMETlab API timeout | Service load, payload, or current quota | Follow live batch limits; retry with backoff and record failures |\n| chemprop training overfits | Random split | Use scaffold split (`--split scaffold_balanced`) |\n| hERG prediction 50\u002F50 | Out-of-distribution | Check applicability domain |\n| QED calculation fails | Molecule is missing, unsanitized, or unsupported | Reject parse failures; sanitize inputs and handle calculation exceptions |\n| ProTox endpoints missing | Web scrape uses CSS selector | Use formal API |\n| BBB+ true but TPSA > 90 | Different BBB model | Use the simple physicochemical screen as an orthogonal heuristic |\n| Predictions inconsistent across runs | Random seed for chemprop ensemble | `--seed 42` and reuse model |\n| Calibration mismatch | DL native probabilities not calibrated | Apply Platt scaling on validation set |\n\n## References\n\n- Fu et al., *Nucleic Acids Res.* 52:W422-W431 (2024) -- ADMETlab 3.0 (DOI 10.1093\u002Fnar\u002Fgkae236).\n- Shan M, Jiang C, Chen J, Qin L-P, Qin J-J, Cheng G. *RSC Adv.* 12:3423-3430 (2022) -- D-MPNN\u002FMOE hERG benchmark (DOI 10.1039\u002FD1RA07956E).\n- Karim A, Lee M, Balle T, Sattar A. *J. Cheminformatics* 13:60 (2021) -- CardioTox-net (DOI 10.1186\u002Fs13321-021-00541-z).\n- Banerjee P, Kemmler E, Dunkel M, Preissner R. *Nucleic Acids Res.* 52:W513-W520 (2024) -- ProTox-3.0 (DOI 10.1093\u002Fnar\u002Fgkae303).\n- McShane SA et al. *J. Cheminformatics* 16:75 (2024) -- conformal prediction for molecular-property models (DOI 10.1186\u002Fs13321-024-00870-9).\n- Heid E et al., *J. Chem. Inf. Model.* 64:9-17 (2024) -- Chemprop redesign (DOI 10.1021\u002Facs.jcim.3c01250).\n- Chemprop documentation, training, descriptors, and uncertainty: https:\u002F\u002Fchemprop.readthedocs.io\u002F\n- ADMET-AI official repository and version notes: https:\u002F\u002Fgithub.com\u002Fswansonk14\u002Fadmet_ai\n- SwissADME Terms of Use: https:\u002F\u002Fwww.swissadme.ch\u002Ftermsofuse.php\n- OECD, \"Principles for the Validation, for Regulatory Purposes, of (Q)SAR Models\" (agreed 2004).\n- OECD, *Guidance Document on the Validation of (Q)SAR Models*, OECD Series on Testing and Assessment No. 69 (2007).\n- OECD, \"(Q)SAR Assessment Framework\" (2023).\n- Capuzzi et al., *J. Chem. Inf. Model.* 57:417 (2017) -- PAINS reality check.\n- Wager et al., *ACS Chem. Neurosci.* 1:435 (2010) -- Pfizer CNS MPO.\n- Bickerton et al., *Nat. Chem.* 4:90 (2012) -- QED.\n\n## Related Skills\n\n- chemoinformatics\u002Fmolecular-descriptors - Compute drug-likeness physchem\n- chemoinformatics\u002Fsubstructure-search - PAINS \u002F BRENK \u002F REOS filter\n- chemoinformatics\u002Fqsar-modeling - Build custom QSAR for in-house data\n- chemoinformatics\u002Fmolecular-standardization - Canonicalize before prediction\n- machine-learning\u002Fbiomarker-discovery - Adjacent ML approaches\n- clinical-databases\u002Fpharmacogenomics - Patient genotype overlay\n","## Version Compatibility\n\nReference examples tested with: RDKit 2024.09+, requests 2.31+, DeepChem 2.8+, chemprop 2.0+ (note major API change from 1.x), admet-ai 1.3+, pandas 2.2+.\n\nBefore using code patterns, verify installed versions match. If versions differ:\n- Python: `pip show \u003Cpackage>` then `help(module.function)` to check signatures\n\nIf code throws ImportError, AttributeError, or TypeError, introspect the installed\npackage and adapt the example to match the actual API rather than retrying.\n\n# ADMET Prediction\n\nPredict absorption, distribution, metabolism, excretion, and toxicity properties of drug candidates. ADMET prediction underpins lead selection and de-risking; calibrated, applicability-domain-aware predictions distinguish a working filter from a costly false-confidence rejection. Modern best practice combines online services (ADMETlab 3.0 with uncertainty estimates), open-source models (chemprop D-MPNN), and rule-based filters (Lipinski \u002F Veber \u002F BBB heuristics) -- each with known failure modes.\n\nFor PAINS \u002F Brenk \u002F structural alerts, see `chemoinformatics\u002Fsubstructure-search`. For QSAR model building from in-house data, see `chemoinformatics\u002Fqsar-modeling`.\n\n## ADMET Model Taxonomy\n\n| Tool | Endpoints | Architecture | Uncertainty | Access | Fails when |\n|------|-----------|--------------|-------------|--------|------------|\n| ADMETlab 3.0 | 119 reported features: 77 prediction models, 34 computed properties, 8 rules | Multi-task DMPNN + descriptors for modeled endpoints | Evidential uncertainty for modeled endpoints | Web service; hosted API documented by the authors | Outside training distribution; metals; macrocycles |\n| ADMET-AI | TDC-derived ADMET tasks; inspect installed model metadata | Chemprop D-MPNN | Inspect version-specific outputs; do not assume calibrated uncertainty | Python package | v2 package predictions differ from the v1 paper\u002Fserver |\n| DeepChem MolNet | Dataset-dependent tasks including Tox21, ToxCast, and ClinTox | Model-dependent | Model-dependent | Python package | Coverage and uncertainty depend on the selected dataset\u002Fmodel |\n| pkCSM | Service-defined ADMET endpoints | Graph signatures + ML | Inspect current service output | Web service | Applicability domain and service contract must be checked |\n| SwissADME | Physchem, pharmacokinetics, drug-likeness, and medchem outputs | Published models and rules | None advertised | Web service (no public API) | Automated access is restricted by its terms |\n| ProTox-3.0 | 61 toxicity models\u002Fendpoints | RF\u002FDNN + fingerprints, similarity, and pharmacophore methods | Confidence score | Web service \u002F sample API | Toxicity only; reports LD50 and toxicity class |\n| ADMETpredictor (Simulations Plus) | ~140 | Proprietary | Per-prediction | Commercial | License cost |\n| FAF-Drugs4 | filters | Rule-based | None | Web | Static rules |\n| chemprop (in-house) | User-defined | D-MPNN ± descriptors | Ensemble and other estimators; optional calibration | Python package | Requires suitable training and calibration data |\n\n**Decision:** For batch screening with no in-house data, **ADMETlab 3.0** provides 119 reported platform features and uncertainty for modeled endpoints through its web service and hosted API; verify the live API documentation before automating access. For a sufficiently large, relevant in-house endpoint dataset, benchmark a **chemprop D-MPNN**, descriptors, and simpler baselines under a deployment-relevant split rather than assuming a universal sample-size threshold. Shan et al. (2022) reported an AUC of 0.956 for a D-MPNN combined with 206 MOE descriptors on their random-split hERG benchmark.\n\n## Decision Tree by Scenario\n\n| Scenario | Workflow | Reasoning |\n|----------|----------|-----------|\n| Library triage, no in-house data | ADMETlab 3.0 API batch | Broad platform coverage plus modeled-endpoint uncertainty |\n| Single endpoint, adequate in-house data | Benchmark chemprop D-MPNN, descriptors, and simpler baselines | Select by prospective or deployment-relevant validation |\n| Need calibrated probabilities | chemprop with ensemble + Platt | Native deep learning rarely calibrated |\n| FDA \u002F regulatory submission | OECD-compliant QSAR with AD | See OECD principles below |\n| Quick annotation for VS | Lipinski, Veber, and QED reported separately | Rank or annotate; do not impose a universal QED gate |\n| BBB penetration | Simple screen: TPSA \u003C= 90, MW \u003C= 500, HBD \u003C= 3 | Repository heuristic; not the six-factor CNS MPO |\n| Cardiotox liability | ADMETlab hERG + ProTox-3.0 cardiotoxicity + literature check | Both hosted endpoints model hERG blockade; compare applicability domains and assay definitions |\n| Drug-drug interaction (CYP) | CYP1A2\u002F2C9\u002F2C19\u002F2D6\u002F3A4 inhibitor + substrate | Standard set of 5 CYPs |\n\n## OECD QSAR Principles (5 Pillars)\n\nFor regulatory-grade ADMET QSAR (REACH, ECHA, FDA submissions), models must satisfy:\n\n1. **Defined endpoint** -- specific bioassay, units, conditions\n2. **Unambiguous algorithm** -- reproducible model + code\n3. **Defined applicability domain (AD)** -- where the model is valid\n4. **Appropriate statistical validation** -- external test set, cross-validation\n5. **Mechanistic interpretation** -- biological \u002F chemical rationale\n\nFor non-regulatory work, AD assessment is still critical. The OECD's *applicability domain* is the workhorse: predictions outside the AD are unreliable, but operational AD measures (leverage, kNN, conformal prediction) often disagree.\n\n## Applicability Domain Methods\n\n| Method | Definition | Flags out-of-AD when |\n|--------|-----------|----------------------|\n| kNN distance | Mean distance to k nearest neighbors in training set | > training-set distribution P95 |\n| Leverage (Williams) | Hat-matrix diagonal | > 3p\u002Fn (p = features, n = compounds) |\n| Density (KDE on PCA) | Density in feature space | \u003C density of training set P5 |\n| Conformal prediction | Per-prediction confidence interval | Interval > tolerance |\n| Bayesian variance | Ensemble or MC-dropout variance | > training-set variance P95 |\n\nFor deep-learning ADMET, **conformal prediction** can provide calibrated prediction sets or intervals when its exchangeability and calibration assumptions are appropriate (McShane et al. 2024).\n\n## ADMETlab 3.0 API\n\nADMETlab 3.0 reports 119 platform features: 77 prediction models, 34 computed physicochemical properties, and 8 medicinal-chemistry rules. The modeled endpoints include prediction uncertainty; do not imply that computed properties and rules have model uncertainty.\n\n**Goal:** Obtain the platform's 119 features for a batch of SMILES, including uncertainty for the 77 modeled endpoints, using the hosted service.\n\n**Approach:** Follow the live ADMETlab 3.0 API tutorial to wash molecules, submit batch predictions, and retrieve the returned results. The 2024 paper documents API\u002Fbatch support and modeled-endpoint uncertainty. Obtain current rate limits, routes, payloads, task identifiers, and output contracts from the live official documentation rather than attributing them to the paper or hard-coding an unofficial example.\n\n```python\nimport pandas as pd\n\n# After submitting with the current official API example, load its CSV output.\nresults = pd.read_csv('admetlab3_results.csv')\n# Preserve the uncertainty columns and task identifier in downstream reports.\n```\n\nADMETlab 3.0 endpoints (sample):\n- Absorption: Caco-2 permeability (logPapp), HIA (%), Pgp inhibitor\u002Fsubstrate, MDCK\n- Distribution: BBB+, PPB (%), VDss (L\u002Fkg), Fu (fraction unbound)\n- Metabolism: CYP1A2\u002F2C9\u002F2C19\u002F2D6\u002F3A4 inhibitor \u002F substrate\n- Excretion: CL (mL\u002Fmin\u002Fkg), T1\u002F2 (h)\n- Toxicity: hERG, AMES, hepatotoxicity (DILI), carcinogenicity, immunotoxicity, mutagenicity, respiratory, skin, eye, cardiotoxicity, mitochondrial, NR-AR, NR-ER, SR-MMP\n- Drug-likeness: Lipinski, Veber, Ghose, Egan, Muegge, QED, SAscore\n\n## chemprop D-MPNN for Custom Endpoints\n\nWhen in-house data is available, train a target-specific model. chemprop provides a widely used open-source D-MPNN architecture with atom\u002Fbond features and optional molecular descriptors; benchmark it against appropriate baselines on the project's data.\n\n**Goal:** Train a target-specific ADMET classifier or regressor on in-house bioassay data.\n\n**Approach:** Use the installed Chemprop 2.x CLI with a scaffold split, a release-supported descriptor featurizer, replicated models, and an explicit prediction-time uncertainty\u002Fcalibration workflow.\n\n```python\n# Chemprop 2.2 CLI; verify flags against the installed release.\n# chemprop train --data-path data.csv --task-type classification \\\n#                --save-dir model_dir --split-type scaffold_balanced \\\n#                --molecule-featurizers v1_rdkit_2d_normalized \\\n#                --num-replicates 5 --ensemble-size 5\n# chemprop predict --test-path test.csv --model-paths model_dir \\\n#                  --uncertainty-method ensemble --preds-path predictions.csv\n\n# Or chemprop 2.x programmatic API (full programmatic API documented at chemprop.readthedocs.io)\n# See chemoinformatics\u002Fqsar-modeling for the full chemprop 2.x training pipeline.\n```\n\n**Key:** Replicates and an ensemble estimator produce an uncertainty estimate, not automatic calibration. Fit and evaluate a documented calibrator on a separate calibration set when calibrated probabilities or intervals are required. Descriptor benefit must be demonstrated on the intended endpoint.\n\n## hERG Cardiotoxicity Endpoint\n\nhERG (KCNH2) blockade can contribute to QT prolongation and Torsades de Pointes and is an important non-clinical cardiac-safety endpoint. Follow the current ICH S7B\u002FE14 and regulator-specific guidance applicable to the program rather than treating one model output as a regulatory conclusion.\n\n| Model | Architecture | Training data | AUC | Reference |\n|-------|--------------|---------------|-----|-----------|\n| Shan et al. D-MPNN + MOE | D-MPNN + 206 MOE descriptors | 7,889 compounds | 0.956 (random split) | Shan 2022 |\n| CardioTox-net | Five DL base representations + neural meta-ensemble | BindingDB, ChEMBL, and literature | 0.930 (10-fold meta-validation) | Karim 2021 |\n| ADMETlab 3.0 hERG | DMPNN multi-task | Internal | 0.92 (reported) | Fu 2024 |\n| ProTox-3.0 cardiotoxicity | RF-based classifier | 5,252 ChEMBL compounds with hERG IC50\u002FKi | 0.86 CV; 0.95 external | Banerjee 2024 |\n\n**Interpretation:** A single-model probability > 0.5 is NOT a kill signal. Triangulate multiple hERG-specific models and a literature search. ProTox-3.0 calls the endpoint cardiotoxicity, but its model specifically predicts small-molecule hERG blockers; it should not be treated as an independent non-hERG mechanism. Consider exposure relative to measured hERG potency and confirm important decisions experimentally rather than applying a universal safe\u002Funsafe IC50 cutoff.\n\n## CYP Inhibition (DDI Risk)\n\n5 CYP isoforms cover most clinically relevant DDIs:\n\n| CYP | Substrates (drugs) | Inhibitor flag if predicted prob | Action |\n|-----|--------------------|-----------------------------------|--------|\n| CYP3A4 | many drug classes | Model-specific threshold | Interpret inhibitor and substrate assays separately |\n| CYP2D6 | beta-blockers, antidepressants | Model-specific threshold | Include polymorphism and exposure context |\n| CYP2C9 | warfarin, NSAIDs | Model-specific threshold | Evaluate clinical substrate\u002Fexposure context |\n| CYP2C19 | PPIs, clopidogrel | Model-specific threshold | Include polymorphism and assay context |\n| CYP1A2 | caffeine, theophylline | Model-specific threshold | Include induction, diet, and smoking context |\n\n## PAINS, BRENK, REOS Filters\n\nADMET prediction is separate from structural alerts; combine. See `chemoinformatics\u002Fsubstructure-search` for PAINS\u002FBRENK\u002FREOS pattern catalogs.\n\n```python\nfrom rdkit.Chem.FilterCatalog import FilterCatalog, FilterCatalogParams\n\ndef alerts(mol, catalogs=('PAINS_A', 'BRENK', 'ZINC')):\n    params = FilterCatalogParams()\n    for cat in catalogs:\n        params.AddCatalog(getattr(FilterCatalogParams.FilterCatalogs, cat))\n    catalog = FilterCatalog(params)\n    hits = catalog.GetMatches(mol)\n    return [h.GetDescription() for h in hits]\n```\n\n## Lipinski \u002F Veber \u002F Drug-Likeness\n\nSee `chemoinformatics\u002Fmolecular-descriptors` for full physchem table. Quick filter:\n\n```python\nfrom rdkit.Chem import Descriptors, Lipinski, QED\n\ndef druglike_score(mol):\n    mw = Descriptors.MolWt(mol)\n    logp = Descriptors.MolLogP(mol)\n    hbd = Lipinski.NumHDonors(mol)\n    hba = Lipinski.NumHAcceptors(mol)\n    tpsa = Descriptors.TPSA(mol)\n    rotbonds = Lipinski.NumRotatableBonds(mol)\n    qed = QED.qed(mol)\n\n    lipinski_violations = sum([mw > 500, logp > 5, hbd > 5, hba > 10])\n    veber_pass = rotbonds \u003C= 10 and tpsa \u003C= 140\n    bbb_simple_screen = tpsa \u003C= 90 and mw \u003C= 500 and hbd \u003C= 3\n\n    return {'MW': mw, 'LogP': logp, 'HBD': hbd, 'HBA': hba,\n            'TPSA': tpsa, 'RotBonds': rotbonds, 'QED': qed,\n            'Lipinski_violations': lipinski_violations,\n            'Veber_pass': veber_pass, 'BBB_simple_screen': bbb_simple_screen}\n```\n\n## Per-Tool Failure Modes\n\n### ADMETlab 3.0 -- out-of-distribution prediction\n\n**Trigger:** Chemistry materially unlike the service's documented training\u002Fapplicability domain, such as metal-containing complexes, many peptides, PROTACs, or unusual macrocycles.\n\n**Mechanism:** ADMETlab training set is drug-like organic molecules. Predictions on PROTACs, macrocycles, peptides extrapolate.\n\n**Symptom:** High reported uncertainty, disagreement with neighbors or orthogonal models, or unstable conclusions under reasonable preprocessing.\n\n**Fix:** Check uncertainty band; if interval is broad, do not trust point estimate. For PROTACs \u002F macrocycles, prefer literature-derived experimental data.\n\n### hERG D-MPNN -- training data bias\n\n**Trigger:** Compound is novel chemotype not in training set (drug-like but in unexplored region).\n\n**Mechanism:** D-MPNN learns local chemical features; for genuinely new scaffolds, extrapolation is unreliable.\n\n**Symptom:** Model predicts hERG- (false negative) for compound that experimentally inhibits.\n\n**Fix:** Use ensemble + applicability-domain assessment (kNN distance, ensemble variance). If kNN distance to training set > P95, treat prediction as low-confidence.\n\n### CYP3A4 inhibitor + substrate ambiguity\n\n**Trigger:** Model trained on either inhibitor OR substrate; predictions confused.\n\n**Mechanism:** CYP3A4 inhibitors and substrates have similar SAR; many compounds are both.\n\n**Symptom:** Both classes report > 0.5.\n\n**Fix:** Two separate models (inhibitor model, substrate model); compounds that score high in both are flagged for in vitro confirmation.\n\n### SwissADME -- no API\n\n**Trigger:** Wanting to batch programmatically.\n\n**Mechanism:** SwissADME's terms restrict automated crawler\u002Fdata-retrieval access, and no public API is documented.\n\n**Symptom:** No programmatic access; manual web upload only.\n\n**Fix:** Use a currently documented programmatic service and follow its access policy; for ADMETlab 3.0, verify the live API tutorial before writing a client.\n\n### PAINS as a kill filter\n\n**Trigger:** Treating PAINS_A match as a categorical exclusion.\n\n**Mechanism:** PAINS is calibrated against HTS assay-interference; matches do NOT predict failed drug development.\n\n**Symptom:** Library purged of valid leads (curcumin analogs, polyphenol natural products).\n\n**Fix:** Flag PAINS for orthogonal-assay confirmation; do not exclude pre-emptively. See substructure-search for details.\n\n### Class-imbalanced AMES dataset\n\n**Trigger:** Training\u002Fpredicting AMES mutagenicity.\n\n**Mechanism:** Public AMES datasets can be imbalanced and differ in assay definition and curation; aggregate accuracy can therefore be misleading.\n\n**Symptom:** Model reports high accuracy but predicts negative for all.\n\n**Fix:** Report class balance and use suitable metrics such as PR-AUC, ROC-AUC, MCC, or balanced accuracy. Compare class weighting or resampling inside training folds without leaking validation\u002Ftest data.\n\n## Reconciliation Across Models\n\nWhen ADMETlab, ProTox-3.0, and an independently trained chemprop model disagree on hERG:\n- All predict hERG+ -> higher concern; plan in vitro patch-clamp\n- Results disagree -> inspect applicability domains, activity thresholds, and assay definitions before deciding\n- All predict hERG- -> lower concern, but still consider in vitro screening for clinical candidates and novel chemotypes\n- Do not count correlated models as independent evidence merely because they are hosted by different services\n\n## Common Errors\n\n| Symptom | Cause | Fix |\n|---------|-------|-----|\n| ADMETlab API timeout | Service load, payload, or current quota | Follow live batch limits; retry with backoff and record failures |\n| chemprop training overfits | Random split | Use scaffold split (`--split scaffold_balanced`) |\n| hERG prediction 50\u002F50 | Out-of-distribution | Check applicability domain |\n| QED calculation fails | Molecule is missing, unsanitized, or unsupported | Reject parse failures; sanitize inputs and handle calculation exceptions |\n| ProTox endpoints missing | Web scrape uses CSS selector | Use formal API |\n| BBB+ true but TPSA > 90 | Different BBB model | Use the simple physicochemical screen as an orthogonal heuristic |\n| Predictions inconsistent across runs | Random seed for chemprop ensemble | `--seed 42` and reuse model |\n| Calibration mismatch | DL native probabilities not calibrated | Apply Platt scaling on validation set |\n\n## References\n\n- Fu et al., *Nucleic Acids Res.* 52:W422-W431 (2024) -- ADMETlab 3.0 (DOI 10.1093\u002Fnar\u002Fgkae236).\n- Shan M, Jiang C, Chen J, Qin L-P, Qin J-J, Cheng G. *RSC Adv.* 12:3423-3430 (2022) -- D-MPNN\u002FMOE hERG benchmark (DOI 10.1039\u002FD1RA07956E).\n- Karim A, Lee M, Balle T, Sattar A. *J. Cheminformatics* 13:60 (2021) -- CardioTox-net (DOI 10.1186\u002Fs13321-021-00541-z).\n- Banerjee P, Kemmler E, Dunkel M, Preissner R. *Nucleic Acids Res.* 52:W513-W520 (2024) -- ProTox-3.0 (DOI 10.1093\u002Fnar\u002Fgkae303).\n- McShane SA et al. *J. Cheminformatics* 16:75 (2024) -- conformal prediction for molecular-property models (DOI 10.1186\u002Fs13321-024-00870-9).\n- Heid E et al., *J. Chem. Inf. Model.* 64:9-17 (2024) -- Chemprop redesign (DOI 10.1021\u002Facs.jcim.3c01250).\n- Chemprop documentation, training, descriptors, and uncertainty: https:\u002F\u002Fchemprop.readthedocs.io\u002F\n- ADMET-AI official repository and version notes: https:\u002F\u002Fgithub.com\u002Fswansonk14\u002Fadmet_ai\n- SwissADME Terms of Use: https:\u002F\u002Fwww.swissadme.ch\u002Ftermsofuse.php\n- OECD, \"Principles for the Validation, for Regulatory Purposes, of (Q)SAR Models\" (agreed 2004).\n- OECD, *Guidance Document on the Validation of (Q)SAR Models*, OECD Series on Testing and Assessment No. 69 (2007).\n- OECD, \"(Q)SAR Assessment Framework\" (2023).\n- Capuzzi et al., *J. Chem. Inf. Model.* 57:417 (2017) -- PAINS reality check.\n- Wager et al., *ACS Chem. Neurosci.* 1:435 (2010) -- Pfizer CNS MPO.\n- Bickerton et al., *Nat. Chem.* 4:90 (2012) -- QED.\n\n## Related Skills\n\n- chemoinformatics\u002Fmolecular-descriptors - Compute drug-likeness physchem\n- chemoinformatics\u002Fsubstructure-search - PAINS \u002F BRENK \u002F REOS filter\n- chemoinformatics\u002Fqsar-modeling - Build custom QSAR for in-house data\n- chemoinformatics\u002Fmolecular-standardization - Canonicalize before prediction\n- machine-learning\u002Fbiomarker-discovery - Adjacent ML approaches\n- clinical-databases\u002Fpharmacogenomics - Patient genotype overlay\n",{"name":5,"source":24,"license":15,"tool_type":25,"description":7,"imported_at":26,"source_path":27,"primary_tool":28,"source_repository":29},"https:\u002F\u002Fgithub.com\u002FGPTomics\u002FbioSkills","python","2026-09-20","chemoinformatics\u002Fadmet-prediction\u002FSKILL.md","ADMETlab","GPTomics\u002FbioSkills","main","2d1750717747097b602acf30bd09a7cbd4b4c6ea","awesome_skills\u002Fbiology\u002Fbio-admet-prediction","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous","2026-11-09 11:45:36",{"source":8,"skillName":5,"commands":37,"prompt":47,"zipAvailable":48},[38,41,44],{"tool":39,"command":40},"npx","npx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill bio-admet-prediction",{"tool":42,"command":43},"bunx","bunx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill bio-admet-prediction",{"tool":45,"command":46},"pnpm","pnpm dlx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill bio-admet-prediction","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-admet-prediction\u002FSKILL.md",true,[50,56],{"path":51,"resourceType":52,"resourceTypeDesc":53,"sizeBytes":54,"sourceHtmlUrl":55},"examples\u002Fpredict_admet.py","SCRIPT","脚本",4024,"https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-admet-prediction\u002Fexamples\u002Fpredict_admet.py",{"path":57,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":60,"sourceHtmlUrl":61},"usage-guide.md","REFERENCE","参考文档",2275,"https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-admet-prediction\u002Fusage-guide.md",[63,71,78,85,92,99],{"skillNo":64,"name":65,"nameZh":6,"description":66,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":67,"categories":68,"keywords":69,"publishedAt":70},"01M7FFFZPC2025JQQHPHDX154J","dei-assessment","Assess diversity, equity, and inclusion with metrics. TRIGGERS - Use when user needs help with dei-assessment related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fdei-assessment\u002FSKILL.md",[],[],"2026-11-16 10:05:34",{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":70},"01M7FFFWN8H4WA7A6Y398QKB6J","dei-action-plan","Create DEI action plans with goals and initiatives. TRIGGERS - Use when user needs help with dei-action-plan related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fdei-action-plan\u002FSKILL.md",[],[],{"skillNo":79,"name":80,"nameZh":6,"description":81,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":70},"01M7FFFSHTM0QZRVZ2NRJTJ0PY","culture-fit-assessment","Design culture add assessment processes. TRIGGERS - Use when user needs help with culture-fit-assessment related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fculture-fit-assessment\u002FSKILL.md",[],[],{"skillNo":86,"name":87,"nameZh":6,"description":88,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":89,"categories":90,"keywords":91,"publishedAt":70},"01M7FFFMA12RKEQA38KHJSXXPS","cowork-to-code-bridge","Use an already-installed, independently verified cowork-to-code bridge to run narrowly approved actions on the user's own macOS, Linux, or WSL2 machine through a local file queue.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcowork-to-code-bridge\u002FSKILL.md",[],[],{"skillNo":93,"name":94,"nameZh":6,"description":95,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":96,"categories":97,"keywords":98,"publishedAt":70},"01M7FFFH1F6M8YRGW5V487GH1K","counter-offer-response","Draft counter-offer responses for retention. TRIGGERS - Use when user needs help with counter-offer-response related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcounter-offer-response\u002FSKILL.md",[],[],{"skillNo":100,"name":101,"nameZh":6,"description":102,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":103,"categories":104,"keywords":105,"publishedAt":70},"01M7FFFDGEHV8Y7J9N40HZXW52","continuous-feedback-system","Design continuous feedback systems. TRIGGERS - Use when user needs help with continuous-feedback-system related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcontinuous-feedback-system\u002FSKILL.md",[],[]]