[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6QHNT267S3VBAGAMY66AFF6":3,"skill-same:01M6QHNT267S3VBAGAMY66AFF6":84},{"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":38,"linkOnly":39,"markdownContent":40,"markdownBody":41,"frontMatter":42,"refName":49,"commitSha":50,"skillDirectory":51,"forksCount":14,"repositoryHtmlUrl":52,"repositoryHomepageUrl":53,"ownerHtmlUrl":54,"ownerBlogUrl":55,"ownerDescription":56,"updatedAt":57,"install":58,"resources":71},"01M6QHNT267S3VBAGAMY66AFF6","scvelo",null,"Performs RNA velocity analysis with scVelo from spliced and unspliced single-cell RNA counts. Fits deterministic or dynamical models, examines gene phase portraits, builds velocity graphs, estimates relative latent time, and ranks velocity-associated genes. Use for directional trajectory hypotheses and kinetic-model diagnostics alongside Scanpy; velocity alone does not establish cell fate or causal drivers.","Tashima-Tarsh\u002FDisha6.6","Tashima-Tarsh","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F269906498?v=4","01M6PBH98X03XERH0HFT4PFB1P","NITISH KUMAR","01M6PBH9DA0M6KM3Z12P848F0K",0,"BSD-3-Clause","https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fkd-scvelo\u002FSKILL.md",[],[5,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37],"velocity","phase","gene","loom","h5ad","scientific","RNA","local","time","diagnostics","deterministic","default","model","stack","credentials","scanpy","analysis","network","pandas","2026-11-10 10:12:30",false,"---\nname: scvelo\ndescription: Performs RNA velocity analysis with scVelo from spliced and unspliced single-cell RNA counts. Fits deterministic or dynamical models, examines gene phase portraits, builds velocity graphs, estimates relative latent time, and ranks velocity-associated genes. Use for directional trajectory hypotheses and kinetic-model diagnostics alongside Scanpy; velocity alone does not establish cell fate or causal drivers.\nlicense: BSD-3-Clause\ncompatibility: Requires Python 3.13 for the tested stack with scvelo 0.3.4, scanpy 1.12.4, anndata 0.13.4, numpy 2.5.3 and pandas 2.3.3. Loom import requires loompy. Local H5AD analysis needs no network or credentials. The default stochastic solver is incompatible with NumPy 2.\nmetadata:\n  version: \"2.0\"\n  skill-author: Kuan-lin Huang\n  last-reviewed: \"2026-10-01\"\ncategory: \"scientific-computing\"\n---\n\n# scVelo RNA Velocity\n\n## When to use\n\nUse for kinetic analysis of aligned spliced\u002Funspliced RNA counts, directional\ntrajectory hypotheses, gene-level phase portraits, and dynamical latent time.\nUse the [model reference](references\u002Fvelocity_models.md) for assumptions,\ntransition probabilities, and an optional CellRank handoff.\n\nRNA velocity estimates an expression derivative under a model. Smooth arrows\nare not observed cell movement, lineage tracing, causal drivers, or proof of\nfuture fate. The dynamical estimator is not automatically more accurate than\na steady-state estimator on every dataset.\n\n## Tested environment and compatibility\n\nThe maintained workflow targets scVelo **0.3.4**, Scanpy **1.12.4**, AnnData\n**0.13.4**, NumPy **2.5.3**, pandas **2.3.3**, SciPy **1.18.1**, Matplotlib\n**3.11.2**, and loompy **3.0.8** on Python 3.13. Use an isolated environment:\n\n```bash\nuv venv --python 3.13 .venv-velocity\nuv pip install --python .venv-velocity\u002Fbin\u002Fpython \\\n  scvelo==0.3.4 scanpy==1.12.4 anndata==0.13.4 numpy==2.5.3 \\\n  pandas==2.3.3 scipy==1.18.1 matplotlib==3.11.2 loompy==3.0.8\n```\n\nThe packages above were resolved from cache and tested in an isolated `uv`\nenvironment; shell paths above use POSIX syntax. Installation requires network\naccess unless packages are cached. No API key is required.\n\n- **Deterministic and dynamical models:** native synthetic fitting, graphs,\n  relative times, plots and H5AD round trips are tested on this stack.\n- **Stochastic model:** scVelo 0.3.4's default GLS scalar assignment fails with\n  NumPy 2. Scanpy 1.12 requires NumPy 2, so installing `numpy\u003C2` beside current\n  Scanpy is not a solution. A separately validated legacy Scanpy\u002FNumPy environment\n  is required; that legacy stack was not executed in this review. The helper\n  rejects this combination before modifying data, without changing the model.\n- **pandas 3:** scVelo 0.3.4 dynamical fitting and some plots use operations no\n  longer supported by pandas 3. Keep the explicit pandas pin.\n- The upstream stable tutorial still contains removed `scv.read` and older\n  preprocessing calls. Check the [0.3.4 source](https:\u002F\u002Fgithub.com\u002Ftheislab\u002Fscvelo\u002Ftree\u002Fv0.3.4)\n  when a tutorial disagrees with the installed API.\n\n## Input contract\n\n1. Obtain spliced and unspliced counts from a velocity-aware quantifier. Keep\n   quantifier\u002Fversion, reference annotation, counting mode, sample IDs and cell\n   barcode mapping. Upstream quantification is outside the bundled script.\n2. Supply **cells by genes** matrices in `adata.layers['spliced']` and\n   `adata.layers['unspliced']`, with identical cell\u002Fgene ordering and unique IDs.\n   Counts must be finite and nonnegative; fractional count estimates are allowed.\n   Never use logged, scaled, residualized, or batch-corrected values as counts.\n   Numerical inspection alone cannot establish that data are raw.\n3. Inspect per-library\u002Fcluster coverage, doublets, ambient RNA, zero-count cells,\n   and annotation compatibility before fitting. Preserve the original full-gene\n   count file. Retained-gene backups do not preserve filtered-out genes.\n4. Resolve barcode prefixes and gene identifiers explicitly when combining\n   files. Do not silently strip library IDs, intersect away most cells, or make\n   duplicated biological IDs unique without understanding why they repeat.\n\nRead H5AD through AnnData. scVelo 0.3.4 has no `scv.read` or `scv.DataFrame`:\n\n```python\nimport anndata as ad\nadata = ad.read_h5ad(\"velocity_counts.h5ad\")\n```\n\nFor legacy loom input use `ad.io.read_loom('counts.loom', X_name='spliced',\nsparse=True)`. AnnData 0.13 deprecates loom; convert to H5AD for further work.\nIts `layers[None]` aliases `X`: do not delete it or iterate all layers as if every\nkey were a string. The helper accesses only named velocity layers. Native loom\nreading and exact, reordered metadata alignment are covered by the tests;\nAnnData 0.13 `write_loom` is not used.\n\n## Run the maintained workflow\n\nThe [bundled script](scripts\u002Frna_velocity_workflow.py) accepts local files and\nnever downloads a demonstration dataset on startup:\n\n```bash\nMPLBACKEND=Agg .venv-velocity\u002Fbin\u002Fpython scripts\u002Frna_velocity_workflow.py \\\n  velocity_counts.h5ad --mode dynamical --groupby clusters \\\n  --n-top-genes 2000 --n-neighbors 30 --n-jobs 1 --output-dir velocity_results\n\n# Import a loom and align its raw counts to an existing annotation file:\nMPLBACKEND=Agg .venv-velocity\u002Fbin\u002Fpython scripts\u002Frna_velocity_workflow.py \\\n  counts.loom --processed-h5ad annotated.h5ad --groupby clusters \\\n  --mode deterministic --output-dir velocity_check\n```\n\nRun these from the skill directory. Input filenames and biological annotations\nare illustrative; the same CLI and functions are tested on small generated\nkinetic fixtures. Omit `--groupby` when annotations are unavailable. Use\n`--no-plots` for analysis without computing UMAP.\n\nThe helper performs these steps:\n\n1. Validate layers, IDs, model compatibility and grouping; refuse previously\n   generated moments\u002Fvelocity so preprocessing cannot silently run twice.\n2. Back up retained raw layers as `spliced_counts` and `unspliced_counts` and\n   rebuild `X` from raw spliced counts. Filter genes and normalize `X` and the\n   two count layers on a linear scale.\n3. Apply `sc.pp.log1p` **only to X**, select highly variable genes with Scanpy,\n   and rebuild PCA and neighbors after subsetting. Reusing stale PCA\u002Fneighbors\n   from an unrelated feature set can silently change the velocity model.\n4. Compute `Ms` and `Mu` with `scv.pp.moments(adata, n_neighbors=None)` from the\n   explicit Scanpy graph. Moments are dense: budget memory for multiple\n   cells-by-genes arrays, not only the sparse input.\n5. For dynamical mode, call `recover_dynamics(var_names='all')` on the selected\n   genes before `velocity(mode='dynamical')`; otherwise fit the selected model.\n   Require usable genes and a nonempty velocity graph.\n6. Compute velocity coherence and velocity pseudotime; add latent time only for\n   dynamical fits. Rank velocity-associated genes only with at least two groups\n   and at least two cells per group. Record\u002Fwarn about nonfinite or constant\n   time\u002Fcoherence outputs in `uns['velocity_workflow']['diagnostics']`; a\n   constant pseudotime does not support a trajectory ordering. Ranking is exploratory.\n7. For plots, recompute UMAP from the rebuilt graph, project velocities, save\n   PNGs directly to the requested directory, then save H5AD with package versions\n   and selected parameters. Existing labels remain annotations, not validated\n   cell identities.\n\nThe Python function mutates its argument in place; pass `adata.copy()` to keep\nthe original object. The CLI writes to `output_dir\u002Fadata_velocity.h5ad`.\nVersion 2 changes the former helper's behavior deliberately: raw layer geometry\nis rebuilt, missing requested labels are errors, and plots\u002Fparallelism can be\ncontrolled explicitly. Runtime depends on cells, genes and fit difficulty;\nthere is no universal 10–30 minute expectation.\n\n## Inspect the evidence before interpretation\n\nAfter a dynamical run (dataset-specific gene selection is illustrative):\n\n```python\nimport pandas as pd\nimport scvelo as scv\n\n# Candidate kinetic genes, not experimentally established drivers.\ncandidates = adata.var['fit_likelihood'].dropna().nlargest(6).index.tolist()\nscv.pl.velocity(adata, var_names=candidates, basis='umap', show=False)\n\n# This ranks group-associated velocities, not causal influence or condition DE.\nscv.tl.rank_velocity_genes(adata, groupby='clusters', min_corr=0.3)\nranked = pd.DataFrame(adata.uns['rank_velocity_genes']['names'])\n\nscv.tl.velocity_confidence(adata)\nscv.pl.scatter(adata, color=['velocity_length', 'velocity_confidence'], show=False)\n```\n\nInspect spliced-versus-unspliced phase portraits, coverage across induction and\nrepression, fitted parameters, failed\u002FNaN fits, and branch-specific kinetics.\nCheck sensitivity to gene set, neighbors, subsampling and model assumptions.\nUse time-course labels, perturbations, lineage tracing or labeling experiments\nas independent directional evidence where available. A plausible UMAP alone\ncannot validate a fit, and tuning until arrows match a desired story is not a\nvalidation strategy.\n\nThere is no general minimum of 2,000 cells, universal unspliced fraction, or\nrule that root cells must have the highest unspliced\u002Fspliced ratio. Coverage of\nrelevant kinetic states and measurement quality matter. Negative velocity can\nrepresent repression or model misspecification; it is not by itself evidence\nthat layers were swapped.\n\n## Output interpretation\n\n| Field | Interpretation |\n| --- | --- |\n| `layers['velocity']` | Model-estimated derivative in processed gene-expression space |\n| `var['velocity_genes']` | Genes selected for the velocity graph; distinct from all HVGs |\n| `layers['Ms']`, `layers['Mu']` | Neighbor-averaged linear-scale spliced\u002Funspliced expression |\n| `layers['fit_t']` | Gene-specific fitted time coordinates, dynamical model only |\n| `var['fit_alpha\u002Fbeta\u002Fgamma']` | Fitted rates on the model's inferred scale; not calibrated physical rates |\n| `var['fit_likelihood']` | Relative model-fit diagnostic; not a posterior probability of biological truth |\n| `uns['velocity_graph']` | Sparse positive cosine correlations for candidate transitions; not row-stochastic |\n| `obsm['velocity_umap']` | Projected vectors, only after embedding computation\u002Fplotting |\n| `obs['velocity_pseudotime']` | Graph-based relative ordering |\n| `obs['latent_time']` | Coupled dynamical ordering; normally scaled 0–1, not elapsed hours |\n| `obs['velocity_length']` | Processed-space vector magnitude; not physical cell speed |\n| `obs['velocity_confidence']` | Neighbor velocity coherence, not calibrated uncertainty |\n\nFor fate probabilities use an explicitly normalized transition kernel and a\nvalidated terminal-state definition; see the optional CellRank example in the\nreference. PAGA requires optional igraph and compatible Scanpy internals; it is\nnot part of the tested core workflow or a substitute for fate inference.\n\n## Troubleshooting\n\n| Problem | Action |\n| --- | --- |\n| Missing or mismatched layers | Revisit quantification and explicit ID alignment; X cannot substitute for unspliced counts |\n| Very few velocity genes | Inspect depth, state coverage and phase portraits before altering thresholds |\n| Smooth but implausible arrows | Check model assumptions, batch geometry and individual genes; compare independent evidence |\n| Nonfinite fit or empty graph | Stop interpretation and investigate degenerate features\u002Fcoverage |\n| Excessive memory or fitting time | Use a justified gene set and `n_jobs=1`; account for dense moments and fit arrays |\n| Stochastic failure on NumPy 2 | Use a separately validated compatible legacy stack or explicitly reconsider the model |\n| pandas `unique` error in fit\u002Fplot | Use the tested pandas 2.3.3 pin |\n\n## Sources and verification scope\n\nReviewed 2026-10-01 against the [released source](https:\u002F\u002Fgithub.com\u002Ftheislab\u002Fscvelo\u002Ftree\u002Fv0.3.4),\n[API](https:\u002F\u002Fscvelo.readthedocs.io\u002Fen\u002Flatest\u002Fapi.html),\n[kinetic-model caveats](https:\u002F\u002Fscvelo.readthedocs.io\u002Fen\u002Flatest\u002Fperspectives\u002FPerspectives.html),\nand [AnnData loom reader](https:\u002F\u002Fanndata.readthedocs.io\u002Fen\u002Fstable\u002Fgenerated\u002Fanndata.io.read_loom.html).\nThe main paper is [Bergen et al., 2020](https:\u002F\u002Fdoi.org\u002F10.1038\u002Fs41587-020-0591-3).\nSynthetic execution tests verify software contracts and file\u002Ffigure production;\nthey do not establish biological accuracy, identifiability or dataset-specific\nparameter recovery. Optional CellRank\u002FPAGA analyses remain source-reviewed,\nillustrative extensions. No authenticated remote service is used.\n","# scVelo RNA Velocity\n\n## When to use\n\nUse for kinetic analysis of aligned spliced\u002Funspliced RNA counts, directional\ntrajectory hypotheses, gene-level phase portraits, and dynamical latent time.\nUse the [model reference](references\u002Fvelocity_models.md) for assumptions,\ntransition probabilities, and an optional CellRank handoff.\n\nRNA velocity estimates an expression derivative under a model. Smooth arrows\nare not observed cell movement, lineage tracing, causal drivers, or proof of\nfuture fate. The dynamical estimator is not automatically more accurate than\na steady-state estimator on every dataset.\n\n## Tested environment and compatibility\n\nThe maintained workflow targets scVelo **0.3.4**, Scanpy **1.12.4**, AnnData\n**0.13.4**, NumPy **2.5.3**, pandas **2.3.3**, SciPy **1.18.1**, Matplotlib\n**3.11.2**, and loompy **3.0.8** on Python 3.13. Use an isolated environment:\n\n```bash\nuv venv --python 3.13 .venv-velocity\nuv pip install --python .venv-velocity\u002Fbin\u002Fpython \\\n  scvelo==0.3.4 scanpy==1.12.4 anndata==0.13.4 numpy==2.5.3 \\\n  pandas==2.3.3 scipy==1.18.1 matplotlib==3.11.2 loompy==3.0.8\n```\n\nThe packages above were resolved from cache and tested in an isolated `uv`\nenvironment; shell paths above use POSIX syntax. Installation requires network\naccess unless packages are cached. No API key is required.\n\n- **Deterministic and dynamical models:** native synthetic fitting, graphs,\n  relative times, plots and H5AD round trips are tested on this stack.\n- **Stochastic model:** scVelo 0.3.4's default GLS scalar assignment fails with\n  NumPy 2. Scanpy 1.12 requires NumPy 2, so installing `numpy\u003C2` beside current\n  Scanpy is not a solution. A separately validated legacy Scanpy\u002FNumPy environment\n  is required; that legacy stack was not executed in this review. The helper\n  rejects this combination before modifying data, without changing the model.\n- **pandas 3:** scVelo 0.3.4 dynamical fitting and some plots use operations no\n  longer supported by pandas 3. Keep the explicit pandas pin.\n- The upstream stable tutorial still contains removed `scv.read` and older\n  preprocessing calls. Check the [0.3.4 source](https:\u002F\u002Fgithub.com\u002Ftheislab\u002Fscvelo\u002Ftree\u002Fv0.3.4)\n  when a tutorial disagrees with the installed API.\n\n## Input contract\n\n1. Obtain spliced and unspliced counts from a velocity-aware quantifier. Keep\n   quantifier\u002Fversion, reference annotation, counting mode, sample IDs and cell\n   barcode mapping. Upstream quantification is outside the bundled script.\n2. Supply **cells by genes** matrices in `adata.layers['spliced']` and\n   `adata.layers['unspliced']`, with identical cell\u002Fgene ordering and unique IDs.\n   Counts must be finite and nonnegative; fractional count estimates are allowed.\n   Never use logged, scaled, residualized, or batch-corrected values as counts.\n   Numerical inspection alone cannot establish that data are raw.\n3. Inspect per-library\u002Fcluster coverage, doublets, ambient RNA, zero-count cells,\n   and annotation compatibility before fitting. Preserve the original full-gene\n   count file. Retained-gene backups do not preserve filtered-out genes.\n4. Resolve barcode prefixes and gene identifiers explicitly when combining\n   files. Do not silently strip library IDs, intersect away most cells, or make\n   duplicated biological IDs unique without understanding why they repeat.\n\nRead H5AD through AnnData. scVelo 0.3.4 has no `scv.read` or `scv.DataFrame`:\n\n```python\nimport anndata as ad\nadata = ad.read_h5ad(\"velocity_counts.h5ad\")\n```\n\nFor legacy loom input use `ad.io.read_loom('counts.loom', X_name='spliced',\nsparse=True)`. AnnData 0.13 deprecates loom; convert to H5AD for further work.\nIts `layers[None]` aliases `X`: do not delete it or iterate all layers as if every\nkey were a string. The helper accesses only named velocity layers. Native loom\nreading and exact, reordered metadata alignment are covered by the tests;\nAnnData 0.13 `write_loom` is not used.\n\n## Run the maintained workflow\n\nThe [bundled script](scripts\u002Frna_velocity_workflow.py) accepts local files and\nnever downloads a demonstration dataset on startup:\n\n```bash\nMPLBACKEND=Agg .venv-velocity\u002Fbin\u002Fpython scripts\u002Frna_velocity_workflow.py \\\n  velocity_counts.h5ad --mode dynamical --groupby clusters \\\n  --n-top-genes 2000 --n-neighbors 30 --n-jobs 1 --output-dir velocity_results\n\n# Import a loom and align its raw counts to an existing annotation file:\nMPLBACKEND=Agg .venv-velocity\u002Fbin\u002Fpython scripts\u002Frna_velocity_workflow.py \\\n  counts.loom --processed-h5ad annotated.h5ad --groupby clusters \\\n  --mode deterministic --output-dir velocity_check\n```\n\nRun these from the skill directory. Input filenames and biological annotations\nare illustrative; the same CLI and functions are tested on small generated\nkinetic fixtures. Omit `--groupby` when annotations are unavailable. Use\n`--no-plots` for analysis without computing UMAP.\n\nThe helper performs these steps:\n\n1. Validate layers, IDs, model compatibility and grouping; refuse previously\n   generated moments\u002Fvelocity so preprocessing cannot silently run twice.\n2. Back up retained raw layers as `spliced_counts` and `unspliced_counts` and\n   rebuild `X` from raw spliced counts. Filter genes and normalize `X` and the\n   two count layers on a linear scale.\n3. Apply `sc.pp.log1p` **only to X**, select highly variable genes with Scanpy,\n   and rebuild PCA and neighbors after subsetting. Reusing stale PCA\u002Fneighbors\n   from an unrelated feature set can silently change the velocity model.\n4. Compute `Ms` and `Mu` with `scv.pp.moments(adata, n_neighbors=None)` from the\n   explicit Scanpy graph. Moments are dense: budget memory for multiple\n   cells-by-genes arrays, not only the sparse input.\n5. For dynamical mode, call `recover_dynamics(var_names='all')` on the selected\n   genes before `velocity(mode='dynamical')`; otherwise fit the selected model.\n   Require usable genes and a nonempty velocity graph.\n6. Compute velocity coherence and velocity pseudotime; add latent time only for\n   dynamical fits. Rank velocity-associated genes only with at least two groups\n   and at least two cells per group. Record\u002Fwarn about nonfinite or constant\n   time\u002Fcoherence outputs in `uns['velocity_workflow']['diagnostics']`; a\n   constant pseudotime does not support a trajectory ordering. Ranking is exploratory.\n7. For plots, recompute UMAP from the rebuilt graph, project velocities, save\n   PNGs directly to the requested directory, then save H5AD with package versions\n   and selected parameters. Existing labels remain annotations, not validated\n   cell identities.\n\nThe Python function mutates its argument in place; pass `adata.copy()` to keep\nthe original object. The CLI writes to `output_dir\u002Fadata_velocity.h5ad`.\nVersion 2 changes the former helper's behavior deliberately: raw layer geometry\nis rebuilt, missing requested labels are errors, and plots\u002Fparallelism can be\ncontrolled explicitly. Runtime depends on cells, genes and fit difficulty;\nthere is no universal 10–30 minute expectation.\n\n## Inspect the evidence before interpretation\n\nAfter a dynamical run (dataset-specific gene selection is illustrative):\n\n```python\nimport pandas as pd\nimport scvelo as scv\n\n# Candidate kinetic genes, not experimentally established drivers.\ncandidates = adata.var['fit_likelihood'].dropna().nlargest(6).index.tolist()\nscv.pl.velocity(adata, var_names=candidates, basis='umap', show=False)\n\n# This ranks group-associated velocities, not causal influence or condition DE.\nscv.tl.rank_velocity_genes(adata, groupby='clusters', min_corr=0.3)\nranked = pd.DataFrame(adata.uns['rank_velocity_genes']['names'])\n\nscv.tl.velocity_confidence(adata)\nscv.pl.scatter(adata, color=['velocity_length', 'velocity_confidence'], show=False)\n```\n\nInspect spliced-versus-unspliced phase portraits, coverage across induction and\nrepression, fitted parameters, failed\u002FNaN fits, and branch-specific kinetics.\nCheck sensitivity to gene set, neighbors, subsampling and model assumptions.\nUse time-course labels, perturbations, lineage tracing or labeling experiments\nas independent directional evidence where available. A plausible UMAP alone\ncannot validate a fit, and tuning until arrows match a desired story is not a\nvalidation strategy.\n\nThere is no general minimum of 2,000 cells, universal unspliced fraction, or\nrule that root cells must have the highest unspliced\u002Fspliced ratio. Coverage of\nrelevant kinetic states and measurement quality matter. Negative velocity can\nrepresent repression or model misspecification; it is not by itself evidence\nthat layers were swapped.\n\n## Output interpretation\n\n| Field | Interpretation |\n| --- | --- |\n| `layers['velocity']` | Model-estimated derivative in processed gene-expression space |\n| `var['velocity_genes']` | Genes selected for the velocity graph; distinct from all HVGs |\n| `layers['Ms']`, `layers['Mu']` | Neighbor-averaged linear-scale spliced\u002Funspliced expression |\n| `layers['fit_t']` | Gene-specific fitted time coordinates, dynamical model only |\n| `var['fit_alpha\u002Fbeta\u002Fgamma']` | Fitted rates on the model's inferred scale; not calibrated physical rates |\n| `var['fit_likelihood']` | Relative model-fit diagnostic; not a posterior probability of biological truth |\n| `uns['velocity_graph']` | Sparse positive cosine correlations for candidate transitions; not row-stochastic |\n| `obsm['velocity_umap']` | Projected vectors, only after embedding computation\u002Fplotting |\n| `obs['velocity_pseudotime']` | Graph-based relative ordering |\n| `obs['latent_time']` | Coupled dynamical ordering; normally scaled 0–1, not elapsed hours |\n| `obs['velocity_length']` | Processed-space vector magnitude; not physical cell speed |\n| `obs['velocity_confidence']` | Neighbor velocity coherence, not calibrated uncertainty |\n\nFor fate probabilities use an explicitly normalized transition kernel and a\nvalidated terminal-state definition; see the optional CellRank example in the\nreference. PAGA requires optional igraph and compatible Scanpy internals; it is\nnot part of the tested core workflow or a substitute for fate inference.\n\n## Troubleshooting\n\n| Problem | Action |\n| --- | --- |\n| Missing or mismatched layers | Revisit quantification and explicit ID alignment; X cannot substitute for unspliced counts |\n| Very few velocity genes | Inspect depth, state coverage and phase portraits before altering thresholds |\n| Smooth but implausible arrows | Check model assumptions, batch geometry and individual genes; compare independent evidence |\n| Nonfinite fit or empty graph | Stop interpretation and investigate degenerate features\u002Fcoverage |\n| Excessive memory or fitting time | Use a justified gene set and `n_jobs=1`; account for dense moments and fit arrays |\n| Stochastic failure on NumPy 2 | Use a separately validated compatible legacy stack or explicitly reconsider the model |\n| pandas `unique` error in fit\u002Fplot | Use the tested pandas 2.3.3 pin |\n\n## Sources and verification scope\n\nReviewed 2026-10-01 against the [released source](https:\u002F\u002Fgithub.com\u002Ftheislab\u002Fscvelo\u002Ftree\u002Fv0.3.4),\n[API](https:\u002F\u002Fscvelo.readthedocs.io\u002Fen\u002Flatest\u002Fapi.html),\n[kinetic-model caveats](https:\u002F\u002Fscvelo.readthedocs.io\u002Fen\u002Flatest\u002Fperspectives\u002FPerspectives.html),\nand [AnnData loom reader](https:\u002F\u002Fanndata.readthedocs.io\u002Fen\u002Fstable\u002Fgenerated\u002Fanndata.io.read_loom.html).\nThe main paper is [Bergen et al., 2020](https:\u002F\u002Fdoi.org\u002F10.1038\u002Fs41587-020-0591-3).\nSynthetic execution tests verify software contracts and file\u002Ffigure production;\nthey do not establish biological accuracy, identifiability or dataset-specific\nparameter recovery. Optional CellRank\u002FPAGA analyses remain source-reviewed,\nillustrative extensions. No authenticated remote service is used.\n",{"name":5,"license":15,"category":43,"metadata":44,"description":7,"compatibility":48},"scientific-computing",{"version":45,"skill-author":46,"last-reviewed":47},"2.0","Kuan-lin Huang","2026-10-01","Requires Python 3.13 for the tested stack with scvelo 0.3.4, scanpy 1.12.4, anndata 0.13.4, numpy 2.5.3 and pandas 2.3.3. Loom import requires loompy. Local H5AD analysis needs no network or credentials. The default stochastic solver is incompatible with NumPy 2.","main","0da4485785ee599388900be50db1635489d53056",".agents\u002Fskills\u002Fkd-scvelo","https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6","https:\u002F\u002Fdisha6.6.thenitishkr.in\u002F","https:\u002F\u002Fgithub.com\u002FTashima-Tarsh","www.thenitishkr","thenitishkr | AI • Cyber-security • Research\r\n","2026-11-08 03:33:25",{"source":8,"skillName":5,"commands":59,"prompt":69,"zipAvailable":70},[60,63,66],{"tool":61,"command":62},"npx","npx skills add Tashima-Tarsh\u002FDisha6.6 --skill scvelo",{"tool":64,"command":65},"bunx","bunx skills add Tashima-Tarsh\u002FDisha6.6 --skill scvelo",{"tool":67,"command":68},"pnpm","pnpm dlx skills add Tashima-Tarsh\u002FDisha6.6 --skill scvelo","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fkd-scvelo\u002FSKILL.md",true,[72,78],{"path":73,"resourceType":74,"resourceTypeDesc":75,"sizeBytes":76,"sourceHtmlUrl":77},"references\u002Fvelocity_models.md","REFERENCE","参考文档",9438,"https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fkd-scvelo\u002Freferences\u002Fvelocity_models.md",{"path":79,"resourceType":80,"resourceTypeDesc":81,"sizeBytes":82,"sourceHtmlUrl":83},"scripts\u002Frna_velocity_workflow.py","SCRIPT","脚本",13207,"https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fkd-scvelo\u002Fscripts\u002Frna_velocity_workflow.py",[85,118,143,170,194,218],{"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":89,"licenseFileUrl":6,"sourceHtmlUrl":90,"categories":91,"keywords":97,"publishedAt":38},"01M6QRD07D4TGNDK6VJ4EHF8BV","youtube-full","Use when the user needs YouTube transcripts, video search, channel browsing, playlist extraction, or content monitoring. Trigger phrases: 'get the transcript for', 'search YouTube for', 'what are the latest videos on', 'list this playlist', 'monitor this channel', or any request involving a YouTube URL, video ID, or @handle. Do NOT use for downloading video or audio files, YouTube engagement data (likes, comments), or private\u002Fage-restricted videos.","MIT","https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fyoutube-full\u002FSKILL.md",[92],{"categoryNo":93,"categoryName":94,"taxonomyType":95,"taxonomyTypeDesc":96},"01JBSEEDCAT000000000000016","图像与多媒体","SCENARIO","使用场景",[98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117],"youtube full","youtube","full","transcript","search","monitoring","data","Video","Audio","list","files","do","get","age","trigger","URL","for","browsing","request","youtube transcripts",{"skillNo":119,"name":120,"nameZh":6,"description":121,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":89,"licenseFileUrl":6,"sourceHtmlUrl":122,"categories":123,"keywords":124,"publishedAt":38},"01M6QRCJ3112VYPA97YG7KGMC2","x-twitter-growth","X\u002FTwitter growth engine for building audience, crafting viral content, and analyzing engagement. Use when the user wants to grow on X\u002FTwitter, write tweets or threads, analyze their X profile, research competitors on X, plan a posting strategy, or optimize engagement. Complements social-content (generic multi-platform) with X-specific depth: algorithm mechanics, thread engineering, reply strategy, profile optimization, and competitive intelligence via web search.","https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fx-twitter-growth\u002FSKILL.md",[],[125,126,127,128,102,129,130,131,132,133,134,135,136,137,138,139,140,141,142,114],"twitter","x twitter growth","growth","research","content","engineering","plan","web search","competitors","analyze","engine","write","strategy","use","platform","grow","posting","competitive",{"skillNo":144,"name":145,"nameZh":6,"description":146,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":89,"licenseFileUrl":6,"sourceHtmlUrl":147,"categories":148,"keywords":152,"publishedAt":38},"01M6QRBN3EHQNJY42RERP7A149","write-a-skill","Create new agent skills with proper structure, progressive disclosure, and bundled resources. Use when user wants to create, write, build, or author a new skill.","https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fwrite-a-skill\u002FSKILL.md",[149],{"categoryNo":150,"categoryName":151,"taxonomyType":95,"taxonomyTypeDesc":96},"01JBSEEDCAT000000000000023","Skill与智能体开发",[153,136,154,155,156,157,158,159,138,160,161,49,162,163,164,165,166,167,168,169],"write a skill","github","create","https","productivity","matt pocock","direct","example","build","new skill","structure","new agent","resources","with","tree","progressive disclosure","author",{"skillNo":171,"name":172,"nameZh":6,"description":173,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":89,"licenseFileUrl":6,"sourceHtmlUrl":174,"categories":175,"keywords":176,"publishedAt":38},"01M6QRB6G8P33PNGAF9Y69EJ9T","wrap-up","Close out a launched Claude Managed Agent — recap every primitive the founder now owns, regenerate the single-file overview page, and suggest the next 1-2 upgrades. Use when the user says \"wrap up\", \"close this out\", \"what do I own now\", \"give me the summary\", \"recap the agent\", or when the orchestrator routes phase=wrap-up. primitives_inventory.py tables everything owned (agent, environment, session, memory, outcome, deployment); overview_page.py regenerates a self-contained .\u002Fmy-agent\u002Fagent-overview.html; upgrade_suggester.py ranks the next moves from recorded deferrals plus standing hardening steps. Companion to run-without-you; the last stop before phase=done.","https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fwrap-up\u002FSKILL.md",[],[177,178,179,180,181,182,183,184,185,138,186,109,187,188,189,190,191,192,193,20],"wrap up","up","wrap","gemini","deployment","run","file","hardening","closeout","stop","upgrade","founder","close","companion","you","tables","directions",{"skillNo":195,"name":196,"nameZh":6,"description":197,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":89,"licenseFileUrl":6,"sourceHtmlUrl":198,"categories":199,"keywords":201,"publishedAt":38},"01M6QRA5AYC1RBFZ39GTWGS3N8","workflow-builder","Design and write deterministic multi-agent workflow scripts (.js files in .claude\u002Fworkflows\u002F) for Claude Code's Workflow tool. Use when a user wants to build, create, author, scaffold, or run a custom Claude Code workflow, orchestrate sub-agents (fan-out, pipeline, loop, judge-panel), or automate a repeatable multi-step task across fresh-context agents.","https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fworkflow-builder\u002FSKILL.md",[200],{"categoryNo":150,"categoryName":151,"taxonomyType":95,"taxonomyTypeDesc":96},[202,203,204,205,206,155,182,207,208,209,210,211,136,212,108,213,214,215,216,217],"workflow","workflow builder","builder","design","workflows","code","orchestrate","pipeline","loop","multi-agent","context","across","fresh","scripts","judge","claude code workflow",{"skillNo":219,"name":220,"nameZh":6,"description":221,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":89,"licenseFileUrl":6,"sourceHtmlUrl":222,"categories":223,"keywords":224,"publishedAt":38},"01M6QR7QMCZESKQA0ERW1E657M","weekly-review","Use when someone wants to run a weekly review, close open loops, audit stalled projects and commitments, get their system back to trusted, restart a lapsed review habit, or says \"\u002Fcs:weekly-review\". Walks David Allen's three-phase loop — GET CLEAR, GET CURRENT, GET CREATIVE — with deterministic scripts that inventory open loops, gate the checklist with named gaps, and score commitment health 0-100.","https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fweekly-review\u002FSKILL.md",[],[225,226,227,228,229,230,231,232,233,210,234,235,236,237,238,239,110,240,241,20],"review","weekly review","weekly","deterministic scripts","reflect","feeds","inventory","open","team","projects","project-management","scanner","conversation","back","project","sprint","cadence"]