[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6QE94F30FVCMG7V9BY5237P":3,"skill-same:01M6QE94F30FVCMG7V9BY5237P":93},{"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":23,"publishedAt":43,"linkOnly":44,"markdownContent":45,"markdownBody":46,"frontMatter":47,"refName":56,"commitSha":57,"skillDirectory":58,"forksCount":14,"repositoryHtmlUrl":59,"repositoryHomepageUrl":60,"ownerHtmlUrl":61,"ownerBlogUrl":62,"ownerDescription":63,"updatedAt":64,"install":65,"resources":78},"01M6QE94F30FVCMG7V9BY5237P","datalad",null,"Retrieve, version, and publish scientific datasets with DataLad and git-annex, and capture computational provenance with datalad run, rerun, and containers-run. Use when cloning or fetching data from OpenNeuro, DANDI, datasets.datalad.org, or any DataLad dataset; when a file in a dataset reads as a broken symlink or a small pointer instead of real data; when an analysis needs a machine-readable record of how each output was produced so it can be re-executed; or when publishing a dataset to siblings such as a GitHub repository plus a storage remote. Also use to decide between DataLad and plain Git for a data-carrying repository.","Tashima-Tarsh\u002FDisha6.6","Tashima-Tarsh","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F269906498?v=4","01M6PBH98X03XERH0HFT4PFB1P","NITISH KUMAR","01M6PBH9DA0M6KM3Z12P848F0K",0,"MIT","https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fkal-datalad\u002FSKILL.md",[18],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000033","数据分析师","PROFESSION","适用职业",[5,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42],"edit","re","Variables","remote","scientific","conda","computational","cloning","uv","component","write","read","decide","versioning","how","open","forge","git","run","2026-11-10 10:12:30",false,"---\nname: datalad\ndescription: Retrieve, version, and publish scientific datasets with DataLad and git-annex, and capture computational provenance with datalad run, rerun, and containers-run. Use when cloning or fetching data from OpenNeuro, DANDI, datasets.datalad.org, or any DataLad dataset; when a file in a dataset reads as a broken symlink or a small pointer instead of real data; when an analysis needs a machine-readable record of how each output was produced so it can be re-executed; or when publishing a dataset to siblings such as a GitHub repository plus a storage remote. Also use to decide between DataLad and plain Git for a data-carrying repository.\nlicense: MIT\ncompatibility: Needs datalad 1.6.x on Python 3.10+, plus git and git-annex 10.x. git-annex is not written in Python but installs as a prebuilt wheel from PyPI (`uv pip install git-annex`), from a system package manager, or from conda-forge. Container-based provenance also needs datalad-container (1.2.x) and Singularity\u002FApptainer or Docker. clone, get, and push need network access; credentialed remotes read secrets from the system keyring or from DATALAD_CREDENTIAL_\u003CNAME>_\u003CCOMPONENT> environment variables.\nallowed-tools: Read Write Edit Bash\nmetadata:\n  version: '1.0'\n  category: data-science-and-ml\n  maintainer: Kalaris Labs\n  contributor: Dylan Pulver\n  tags: research data management, data versioning, provenance, reproducibility, open science\n---\n\n# DataLad\n\n## Overview\n\nDataLad is a data management layer over Git and git-annex. Git tracks the dataset\nstructure, small text files, and the history. git-annex tracks the *content* of large\nfiles, storing each file as a key and keeping the bytes somewhere that is not necessarily\nthe local repository.\n\nThat split is the single most important thing to internalise, because it means a freshly\ncloned dataset contains the full history and the full file listing while containing almost\nnone of the data. A 100 TB dataset clones in seconds and occupies a few megabytes. The\nbytes arrive only when asked for, per file, with `datalad get`.\n\nThe second thing DataLad adds is provenance. `datalad run` executes a command and commits\nthe result together with a machine-readable record of the command, its inputs, and its\noutputs. `datalad rerun` reads that record back and re-executes it. This turns \"how was\nthis figure produced\" from an archaeology problem into a command.\n\n## When to use DataLad instead of plain Git\n\nUse DataLad when any of the following holds:\n\n- Files are too large for Git to handle comfortably, or the total exceeds what every\n  collaborator wants on disk.\n- Data lives in more than one place (a lab server, a cluster scratch, S3, a supercomputer)\n  and you need to know which copies exist.\n- The analysis must be re-executable, and a plain commit message is not enough evidence.\n- You are consuming published datasets from OpenNeuro, DANDI, or `datasets.datalad.org`,\n  which are distributed as DataLad datasets.\n- The project nests other datasets inside it and you want each one to keep its own\n  independent history.\n\nUse plain Git when the repository is code and text only, everything fits comfortably in\nGit, and nobody needs partial checkouts. DataLad on top of a small pure-code repository\nadds indirection without buying anything.\n\n## Installation\n\n```bash\n# git-annex is NOT written in Python but is available from PyPI if you already\n# have git itself installed:\nuv pip install git-annex\n# You can also install it first from the system\n# (Debian\u002FUbuntu: apt install git-annex; macOS: brew install git-annex;\n#  conda-forge: conda install -c conda-forge git-annex)\nuv pip install datalad\nuv pip install datalad-container   # only for containers-run\n\ndatalad wtf --section dependencies   # confirm git-annex version is visible\n```\n\nThe PyPI `git-annex` package ships the prebuilt binary as a wheel for Linux, macOS, and\nWindows rather than building the Haskell sources, so it installs like any other Python\ndependency and can be pinned in the same environment as DataLad. It does not bring git\nalong with it.\n\n`datalad wtf` prints the resolved environment and is the first thing to run when behaviour\nlooks impossible. An old or missing git-annex is behind a large share of confusing errors.\n\nDataLad itself is MIT licensed. git-annex is a separate tool under the AGPL, which matters\nonly if you redistribute a modified git-annex rather than call it.\n\n## The failure that bites first: pointers are not data\n\nAfter `datalad clone`, annexed files exist as symlinks into `.git\u002Fannex\u002Fobjects\u002F` (or as\nsmall pointer files where symlinks are unavailable, such as on Windows or a crippled\nfilesystem). Nothing has downloaded the content yet.\n\n```bash\ndatalad clone https:\u002F\u002Fgithub.com\u002FOpenNeuroDatasets\u002Fds000001.git\ncd ds000001\nls sub-01\u002Fanat\u002F            # the file is listed\npython -c \"import nibabel; nibabel.load('sub-01\u002Fanat\u002Fsub-01_T1w.nii.gz')\"   # fails\ndatalad get sub-01\u002Fanat\u002Fsub-01_T1w.nii.gz                                   # now it works\n```\n\nThe failure mode to recognise: a tool reports the file as empty, truncated, corrupt, \"not\na gzip file\", or a broken symlink, and the file size on disk is a few hundred bytes. That\nis a pointer, not a corrupted download. **Run `datalad get` before reading data, and treat\n\"file exists\" as insufficient evidence that its content is present.**\n\nBefore an analysis touches a directory, fetch it explicitly:\n\n```bash\ndatalad get sub-01\u002F                  # everything under a path\ndatalad get -r .                     # everything, including subdatasets\ndatalad get -n -r .                  # subdataset structure only, no file content\n```\n\n`datalad status --annex` reports how much content is present locally, and\n`git annex whereis \u003Cpath>` reports which repositories hold a given file. `whereis` reads\nrecorded state and does not contact the remotes, so it tells you what git-annex last\nlearned rather than what is true right now.\n\nSee [data-access.md](references\u002Fdata-access.md) for finding datasets, subdataset\nbehaviour, dropping content safely, and repairing a dataset.\n\n## Recording provenance with datalad run\n\n`datalad run` is the reason to reach for DataLad in a methods context. It saves the\ncommand alongside its effect, in the same commit:\n\n```bash\ndatalad run -m \"extract brain mask\" \\\n  --input \"sub-01\u002Fanat\u002Fsub-01_T1w.nii.gz\" \\\n  --output \"derivatives\u002Fsub-01_brain.nii.gz\" \\\n  \"bet {inputs} {outputs} -m\"\n```\n\nWhat each part does, and why skipping it hurts:\n\n- `--input` retrieves the content before running, so the command does not fail on a\n  pointer. It also records the dependency, which is what lets `rerun` fetch the same\n  inputs on a different machine.\n- `--output` unlocks or removes the target first, so git-annex does not refuse to write\n  over content it is protecting. Without it, a second run of the same command commonly\n  fails with a permission error on an annexed file that looks read-only.\n- `{inputs}` and `{outputs}` expand to those values. `{pwd}`, `{dspath}`, and `{tmpdir}`\n  are also available, and `{inputs[0]}` indexes individual entries.\n- The commit message carries a JSON run record between `=== Do not change lines below ===`\n  and `^^^ Do not change lines above ^^^`. Do not hand-edit that block; `rerun` parses it.\n\n`datalad run` refuses to start when the dataset has unsaved modifications, because an\nunclean starting state makes the record unreliable. Save or discard first, or pass\n`--explicit` to declare that the listed inputs and outputs are the complete story. Check a\ncommand before committing to it with `--dry-run basic` or `--dry-run command`.\n\nA run that changes nothing produces no commit, exactly as `datalad save` does.\n\n### Re-executing\n\n```bash\ndatalad rerun                       # redo the run recorded at HEAD\ndatalad rerun --report              # show what would be done, change nothing\ndatalad rerun --script recompute.sh # extract the commands instead of running them\ndatalad rerun --since \u003Ccommit> -b check \u003Crevision>   # replay a range onto a new branch\n```\n\nRerunning onto a branch (`-b`) is the safe way to test reproducibility: the replay lands\nsomewhere else, and a diff against the original branch answers whether the outputs came\nback identical.\n\n### Containers\n\nWith the `datalad-container` extension, register an image once and every subsequent run\nrecords which image produced the outputs:\n\n```bash\ndatalad containers-add fsl --url docker:\u002F\u002Fbrainlife\u002Ffsl:6.0.4\ndatalad containers-run -n fsl -m \"brain mask in container\" \\\n  --input \"sub-01\u002Fanat\u002Fsub-01_T1w.nii.gz\" \\\n  --output \"derivatives\u002Fsub-01_brain.nii.gz\" \\\n  \"bet {inputs} {outputs} -m\"\n```\n\nThe image itself is tracked in the dataset, so the software environment travels with the\ndata and the provenance record rather than living in someone's shell history. When only\none container is configured, `-n` may be omitted.\n\nSee [provenance.md](references\u002Fprovenance.md) for the STAMPED principles and the YODA\nproject layout, the run record format, `--explicit` and `--assume-ready` semantics, and\nexporting provenance toward W3C PROV.\n\n## Saving and inspecting changes\n\n```bash\ndatalad status                 # what changed, including subdataset state\ndatalad save -m \"add QC report\" path\u002Fto\u002Ffile\ndatalad save -m \"checkpoint\" -r                 # recurse into subdatasets\ndatalad save -m \"small text file\" --to-git notes.md\n```\n\n`datalad save` decides per file whether content goes to Git or to git-annex, following the\ndataset's `.gitattributes`. Force a file into Git with `--to-git`, which is the right call\nfor code and small text files that should stay directly readable. The `yoda` procedure\n(`datalad create -c yoda`) sets this up for `code\u002F`, `README.md`, and `CHANGELOG.md`\nautomatically.\n\n## Creating a dataset\n\n```bash\ndatalad create my_dataset               # plain dataset\ndatalad create -c yoda my_analysis      # analysis layout (code\u002F tracked in Git,\n                                        # README.md and CHANGELOG.md preconfigured)\ndatalad create -d . inputs\u002Fraw          # register a new subdataset under an existing one\n```\n\n`-c yoda` applies the analysis project layout described in\n[provenance.md](references\u002Fprovenance.md). `-d .` is what registers a new dataset as a\nsubdataset of the parent rather than leaving an unrelated repository inside it.\n\n## Publishing\n\nA DataLad dataset is usually published to two places at once: a Git hosting service for\nthe history, and a storage remote for the annexed content.\n\n```bash\ndatalad create-sibling-github myaccount\u002Fmydataset\ngit annex initremote store type=S3 bucket=my-bucket encryption=none autoenable=true\ndatalad siblings configure -s github --publish-depends store\ndatalad push --to github\n```\n\nThe Git sibling and the storage sibling are created by different tools on purpose. A Git\nsibling is a Git remote, and `datalad create-sibling-*` handles the hosting-service ones.\nAn S3 bucket (or WebDAV, or an SSH directory) is a *git-annex special remote*, not a Git\nremote, so it is created with `git annex initremote`. `datalad siblings` picks the special\nremote up afterwards and treats it like any other. Using `datalad siblings add --url\ns3:\u002F\u002F...` here is the mistake this section exists to prevent: `--url` is a Git remote URL,\nS3 is not, and the `push --to github` below then fails on the `--publish-depends` hop.\n\n`--publish-depends` is what stops the common broken publication: a Git repository whose\nhistory references content that was never uploaded, so collaborators clone successfully\nand then find every `datalad get` failing. Declaring the dependency makes the storage\nsibling publish first, every time.\n\n`datalad push` sends both the Git history and, by default (`--data auto-if-wanted`), the\nannexed content the target is configured to want. Pass `--data anything` to push all\ncontent regardless of the target's preferences.\n\nSee [publishing.md](references\u002Fpublishing.md) for RIA stores, special remotes, credential\nhandling, and configuring which sibling holds what.\n\n## Freeing disk space\n\n```bash\ngit annex whereis sub-01\u002F                 # confirm another copy exists first\ndatalad drop sub-01\u002F                      # remove local content, keep the pointer\ndatalad drop --what all --reckless kill \u003Cpath>   # last resort, destroys data\n```\n\n`datalad drop` refuses by default when it cannot verify another copy of the content\nexists, which is a safety check rather than an obstacle. `--nocheck` and `--if-dirty` are\ndeprecated; the current spelling is `--reckless availability`, and it means what it says.\n`--what` selects between `filecontent` (the default), `allkeys`, `datasets`, and `all`.\n\n## Failure modes worth knowing\n\n| Symptom | Cause | Fix |\n|---|---|---|\n| File reads as empty, truncated, or a broken symlink | Content not retrieved; only the pointer is present | `datalad get \u003Cpath>` |\n| \"Permission denied\" writing an existing output | git-annex write-protects annexed content | Declare it with `--output`, or `datalad unlock \u003Cpath>` |\n| `datalad run` refuses to start | Dataset has unsaved changes | `datalad save` first, or pass `--explicit` |\n| `datalad drop` refuses | No verified second copy of the content | Push to a sibling first, or accept `--reckless availability` |\n| Collaborator clones but every `get` fails | History published without the content | Publish the storage sibling, and set `--publish-depends` |\n| Clone succeeds, subdataset directories are empty | Subdatasets are not installed by default | `datalad get -n -r .`, then `get` the paths you need |\n| Commands behave impossibly | git-annex missing or too old | `datalad wtf --section dependencies` |\n\n## Detailed references\n\n- [data-access.md](references\u002Fdata-access.md): finding published datasets\n  (`registry.datalad.org`, OpenNeuro, DANDI, `datasets.datalad.org` and the `\u002F\u002F\u002F`\n  shortcut), clone and get options, subdataset handling, annex content states, dropping\n  and removing, and `fsck` repair.\n- [provenance.md](references\u002Fprovenance.md): the STAMPED principles and the YODA layout,\n  the run record format, `run` and `rerun` options in full, `containers-run`, and the\n  current state of exporting DataLad provenance toward W3C PROV.\n- [publishing.md](references\u002Fpublishing.md): siblings and their actions,\n  `create-sibling-*` variants, RIA stores, special remotes, `push` semantics, and\n  credential handling.\n\n## Related skills\n\nThe `bids` skill covers the Brain Imaging Data Structure that most of the neuroimaging\ndatasets distributed through DataLad are organised in. A typical workflow clones a BIDS\ndataset with DataLad, validates it with the BIDS tooling, then runs a BIDS-App under\n`datalad containers-run` so the derivatives carry provenance.\n\n## Primary sources\n\n- DataLad documentation: \u003Chttps:\u002F\u002Fdocs.datalad.org\u002Fen\u002Fstable\u002F>\n- DataLad Handbook: \u003Chttps:\u002F\u002Fhandbook.datalad.org\u002Fen\u002Flatest\u002F>\n- `datalad run` chapter: \u003Chttps:\u002F\u002Fhandbook.datalad.org\u002Fen\u002Flatest\u002Fbasics\u002F101-108-run.html>\n- YODA principles: \u003Chttps:\u002F\u002Fhandbook.datalad.org\u002Fen\u002Flatest\u002Fbasics\u002F101-127-yoda.html>\n- STAMPED principles (operationalized from YODA): \u003Chttps:\u002F\u002Fstamped-principles.org>\n- datalad-container: \u003Chttps:\u002F\u002Fdocs.datalad.org\u002Fprojects\u002Fcontainer\u002Fen\u002Fstable\u002F>\n- git-annex: \u003Chttps:\u002F\u002Fgit-annex.branchable.com\u002F>\n- Dataset registry: \u003Chttps:\u002F\u002Fregistry.datalad.org>\n\n## Acknowledgment\n\nTopic scope for this skill was informed in part by @bcmcpher's MIT-licensed\n[datalad-cli](https:\u002F\u002Fgithub.com\u002Fbcmcpher\u002Fmy-skills\u002Ftree\u002Fmain\u002Fplugins\u002Fdatalad-cli)\nplugin (nineteen per-command slash-command skills). The text here is written\nindependently and grounded in the upstream DataLad documentation; overlap is unavoidable\nbecause both cover DataLad, but the structure, style, and specific technical claims are\ndifferent.\n\n## Agent operating procedure\n\n1. **Check the environment.** Confirm the Python environment and library versions (`python -c \"import pkg; print(pkg.__version__)\"`) and inspect the data's shape, types and missing values.\n2. **Pin down the inputs.** Confirm formats, identifiers and parameters from the data or the user. Ask rather than guess any value that changes the result.\n3. **Run a small version first.** Run on a sample or a single fold first and check runtime and memory.\n4. **Execute the full task** using the instructions and references above.\n5. **Validate the result.** Use held-out data, fixed random seeds and appropriate metrics; check for leakage; report uncertainty (CIs, std over seeds).\n6. **Report.** State what was run (versions, commands, parameters), what was checked, and what is still uncertain.\n\n| If this happens | Do this |\n|---|---|\n| Out-of-memory or very slow execution | Subsample, use chunked or lazy computation, or reduce model size, and tell the user what changed. |\n| A function, flag or endpoint in these instructions is missing in the installed version | Check the installed version's own documentation (`help()`, `--help`, official docs), adapt, and tell the user. Never invent an API. |\n| A required input, identifier or parameter is ambiguous | Ask the user, or state the assumption explicitly before running. |\n\n**Integrity rules**\n\n- Never fabricate results, parameters, identifiers, citations or statistics. If something cannot be run or verified, say so plainly.\n- Never report a metric you did not compute in this session; show the code path that produced every number.\n- Treat version-specific details here as possibly outdated: confirm them against the official documentation for the installed version.\n- Ask before actions that cost money, consume shared GPUs or cloud quota, touch personal or patient data, or cannot be undone.\n","# DataLad\n\n## Overview\n\nDataLad is a data management layer over Git and git-annex. Git tracks the dataset\nstructure, small text files, and the history. git-annex tracks the *content* of large\nfiles, storing each file as a key and keeping the bytes somewhere that is not necessarily\nthe local repository.\n\nThat split is the single most important thing to internalise, because it means a freshly\ncloned dataset contains the full history and the full file listing while containing almost\nnone of the data. A 100 TB dataset clones in seconds and occupies a few megabytes. The\nbytes arrive only when asked for, per file, with `datalad get`.\n\nThe second thing DataLad adds is provenance. `datalad run` executes a command and commits\nthe result together with a machine-readable record of the command, its inputs, and its\noutputs. `datalad rerun` reads that record back and re-executes it. This turns \"how was\nthis figure produced\" from an archaeology problem into a command.\n\n## When to use DataLad instead of plain Git\n\nUse DataLad when any of the following holds:\n\n- Files are too large for Git to handle comfortably, or the total exceeds what every\n  collaborator wants on disk.\n- Data lives in more than one place (a lab server, a cluster scratch, S3, a supercomputer)\n  and you need to know which copies exist.\n- The analysis must be re-executable, and a plain commit message is not enough evidence.\n- You are consuming published datasets from OpenNeuro, DANDI, or `datasets.datalad.org`,\n  which are distributed as DataLad datasets.\n- The project nests other datasets inside it and you want each one to keep its own\n  independent history.\n\nUse plain Git when the repository is code and text only, everything fits comfortably in\nGit, and nobody needs partial checkouts. DataLad on top of a small pure-code repository\nadds indirection without buying anything.\n\n## Installation\n\n```bash\n# git-annex is NOT written in Python but is available from PyPI if you already\n# have git itself installed:\nuv pip install git-annex\n# You can also install it first from the system\n# (Debian\u002FUbuntu: apt install git-annex; macOS: brew install git-annex;\n#  conda-forge: conda install -c conda-forge git-annex)\nuv pip install datalad\nuv pip install datalad-container   # only for containers-run\n\ndatalad wtf --section dependencies   # confirm git-annex version is visible\n```\n\nThe PyPI `git-annex` package ships the prebuilt binary as a wheel for Linux, macOS, and\nWindows rather than building the Haskell sources, so it installs like any other Python\ndependency and can be pinned in the same environment as DataLad. It does not bring git\nalong with it.\n\n`datalad wtf` prints the resolved environment and is the first thing to run when behaviour\nlooks impossible. An old or missing git-annex is behind a large share of confusing errors.\n\nDataLad itself is MIT licensed. git-annex is a separate tool under the AGPL, which matters\nonly if you redistribute a modified git-annex rather than call it.\n\n## The failure that bites first: pointers are not data\n\nAfter `datalad clone`, annexed files exist as symlinks into `.git\u002Fannex\u002Fobjects\u002F` (or as\nsmall pointer files where symlinks are unavailable, such as on Windows or a crippled\nfilesystem). Nothing has downloaded the content yet.\n\n```bash\ndatalad clone https:\u002F\u002Fgithub.com\u002FOpenNeuroDatasets\u002Fds000001.git\ncd ds000001\nls sub-01\u002Fanat\u002F            # the file is listed\npython -c \"import nibabel; nibabel.load('sub-01\u002Fanat\u002Fsub-01_T1w.nii.gz')\"   # fails\ndatalad get sub-01\u002Fanat\u002Fsub-01_T1w.nii.gz                                   # now it works\n```\n\nThe failure mode to recognise: a tool reports the file as empty, truncated, corrupt, \"not\na gzip file\", or a broken symlink, and the file size on disk is a few hundred bytes. That\nis a pointer, not a corrupted download. **Run `datalad get` before reading data, and treat\n\"file exists\" as insufficient evidence that its content is present.**\n\nBefore an analysis touches a directory, fetch it explicitly:\n\n```bash\ndatalad get sub-01\u002F                  # everything under a path\ndatalad get -r .                     # everything, including subdatasets\ndatalad get -n -r .                  # subdataset structure only, no file content\n```\n\n`datalad status --annex` reports how much content is present locally, and\n`git annex whereis \u003Cpath>` reports which repositories hold a given file. `whereis` reads\nrecorded state and does not contact the remotes, so it tells you what git-annex last\nlearned rather than what is true right now.\n\nSee [data-access.md](references\u002Fdata-access.md) for finding datasets, subdataset\nbehaviour, dropping content safely, and repairing a dataset.\n\n## Recording provenance with datalad run\n\n`datalad run` is the reason to reach for DataLad in a methods context. It saves the\ncommand alongside its effect, in the same commit:\n\n```bash\ndatalad run -m \"extract brain mask\" \\\n  --input \"sub-01\u002Fanat\u002Fsub-01_T1w.nii.gz\" \\\n  --output \"derivatives\u002Fsub-01_brain.nii.gz\" \\\n  \"bet {inputs} {outputs} -m\"\n```\n\nWhat each part does, and why skipping it hurts:\n\n- `--input` retrieves the content before running, so the command does not fail on a\n  pointer. It also records the dependency, which is what lets `rerun` fetch the same\n  inputs on a different machine.\n- `--output` unlocks or removes the target first, so git-annex does not refuse to write\n  over content it is protecting. Without it, a second run of the same command commonly\n  fails with a permission error on an annexed file that looks read-only.\n- `{inputs}` and `{outputs}` expand to those values. `{pwd}`, `{dspath}`, and `{tmpdir}`\n  are also available, and `{inputs[0]}` indexes individual entries.\n- The commit message carries a JSON run record between `=== Do not change lines below ===`\n  and `^^^ Do not change lines above ^^^`. Do not hand-edit that block; `rerun` parses it.\n\n`datalad run` refuses to start when the dataset has unsaved modifications, because an\nunclean starting state makes the record unreliable. Save or discard first, or pass\n`--explicit` to declare that the listed inputs and outputs are the complete story. Check a\ncommand before committing to it with `--dry-run basic` or `--dry-run command`.\n\nA run that changes nothing produces no commit, exactly as `datalad save` does.\n\n### Re-executing\n\n```bash\ndatalad rerun                       # redo the run recorded at HEAD\ndatalad rerun --report              # show what would be done, change nothing\ndatalad rerun --script recompute.sh # extract the commands instead of running them\ndatalad rerun --since \u003Ccommit> -b check \u003Crevision>   # replay a range onto a new branch\n```\n\nRerunning onto a branch (`-b`) is the safe way to test reproducibility: the replay lands\nsomewhere else, and a diff against the original branch answers whether the outputs came\nback identical.\n\n### Containers\n\nWith the `datalad-container` extension, register an image once and every subsequent run\nrecords which image produced the outputs:\n\n```bash\ndatalad containers-add fsl --url docker:\u002F\u002Fbrainlife\u002Ffsl:6.0.4\ndatalad containers-run -n fsl -m \"brain mask in container\" \\\n  --input \"sub-01\u002Fanat\u002Fsub-01_T1w.nii.gz\" \\\n  --output \"derivatives\u002Fsub-01_brain.nii.gz\" \\\n  \"bet {inputs} {outputs} -m\"\n```\n\nThe image itself is tracked in the dataset, so the software environment travels with the\ndata and the provenance record rather than living in someone's shell history. When only\none container is configured, `-n` may be omitted.\n\nSee [provenance.md](references\u002Fprovenance.md) for the STAMPED principles and the YODA\nproject layout, the run record format, `--explicit` and `--assume-ready` semantics, and\nexporting provenance toward W3C PROV.\n\n## Saving and inspecting changes\n\n```bash\ndatalad status                 # what changed, including subdataset state\ndatalad save -m \"add QC report\" path\u002Fto\u002Ffile\ndatalad save -m \"checkpoint\" -r                 # recurse into subdatasets\ndatalad save -m \"small text file\" --to-git notes.md\n```\n\n`datalad save` decides per file whether content goes to Git or to git-annex, following the\ndataset's `.gitattributes`. Force a file into Git with `--to-git`, which is the right call\nfor code and small text files that should stay directly readable. The `yoda` procedure\n(`datalad create -c yoda`) sets this up for `code\u002F`, `README.md`, and `CHANGELOG.md`\nautomatically.\n\n## Creating a dataset\n\n```bash\ndatalad create my_dataset               # plain dataset\ndatalad create -c yoda my_analysis      # analysis layout (code\u002F tracked in Git,\n                                        # README.md and CHANGELOG.md preconfigured)\ndatalad create -d . inputs\u002Fraw          # register a new subdataset under an existing one\n```\n\n`-c yoda` applies the analysis project layout described in\n[provenance.md](references\u002Fprovenance.md). `-d .` is what registers a new dataset as a\nsubdataset of the parent rather than leaving an unrelated repository inside it.\n\n## Publishing\n\nA DataLad dataset is usually published to two places at once: a Git hosting service for\nthe history, and a storage remote for the annexed content.\n\n```bash\ndatalad create-sibling-github myaccount\u002Fmydataset\ngit annex initremote store type=S3 bucket=my-bucket encryption=none autoenable=true\ndatalad siblings configure -s github --publish-depends store\ndatalad push --to github\n```\n\nThe Git sibling and the storage sibling are created by different tools on purpose. A Git\nsibling is a Git remote, and `datalad create-sibling-*` handles the hosting-service ones.\nAn S3 bucket (or WebDAV, or an SSH directory) is a *git-annex special remote*, not a Git\nremote, so it is created with `git annex initremote`. `datalad siblings` picks the special\nremote up afterwards and treats it like any other. Using `datalad siblings add --url\ns3:\u002F\u002F...` here is the mistake this section exists to prevent: `--url` is a Git remote URL,\nS3 is not, and the `push --to github` below then fails on the `--publish-depends` hop.\n\n`--publish-depends` is what stops the common broken publication: a Git repository whose\nhistory references content that was never uploaded, so collaborators clone successfully\nand then find every `datalad get` failing. Declaring the dependency makes the storage\nsibling publish first, every time.\n\n`datalad push` sends both the Git history and, by default (`--data auto-if-wanted`), the\nannexed content the target is configured to want. Pass `--data anything` to push all\ncontent regardless of the target's preferences.\n\nSee [publishing.md](references\u002Fpublishing.md) for RIA stores, special remotes, credential\nhandling, and configuring which sibling holds what.\n\n## Freeing disk space\n\n```bash\ngit annex whereis sub-01\u002F                 # confirm another copy exists first\ndatalad drop sub-01\u002F                      # remove local content, keep the pointer\ndatalad drop --what all --reckless kill \u003Cpath>   # last resort, destroys data\n```\n\n`datalad drop` refuses by default when it cannot verify another copy of the content\nexists, which is a safety check rather than an obstacle. `--nocheck` and `--if-dirty` are\ndeprecated; the current spelling is `--reckless availability`, and it means what it says.\n`--what` selects between `filecontent` (the default), `allkeys`, `datasets`, and `all`.\n\n## Failure modes worth knowing\n\n| Symptom | Cause | Fix |\n|---|---|---|\n| File reads as empty, truncated, or a broken symlink | Content not retrieved; only the pointer is present | `datalad get \u003Cpath>` |\n| \"Permission denied\" writing an existing output | git-annex write-protects annexed content | Declare it with `--output`, or `datalad unlock \u003Cpath>` |\n| `datalad run` refuses to start | Dataset has unsaved changes | `datalad save` first, or pass `--explicit` |\n| `datalad drop` refuses | No verified second copy of the content | Push to a sibling first, or accept `--reckless availability` |\n| Collaborator clones but every `get` fails | History published without the content | Publish the storage sibling, and set `--publish-depends` |\n| Clone succeeds, subdataset directories are empty | Subdatasets are not installed by default | `datalad get -n -r .`, then `get` the paths you need |\n| Commands behave impossibly | git-annex missing or too old | `datalad wtf --section dependencies` |\n\n## Detailed references\n\n- [data-access.md](references\u002Fdata-access.md): finding published datasets\n  (`registry.datalad.org`, OpenNeuro, DANDI, `datasets.datalad.org` and the `\u002F\u002F\u002F`\n  shortcut), clone and get options, subdataset handling, annex content states, dropping\n  and removing, and `fsck` repair.\n- [provenance.md](references\u002Fprovenance.md): the STAMPED principles and the YODA layout,\n  the run record format, `run` and `rerun` options in full, `containers-run`, and the\n  current state of exporting DataLad provenance toward W3C PROV.\n- [publishing.md](references\u002Fpublishing.md): siblings and their actions,\n  `create-sibling-*` variants, RIA stores, special remotes, `push` semantics, and\n  credential handling.\n\n## Related skills\n\nThe `bids` skill covers the Brain Imaging Data Structure that most of the neuroimaging\ndatasets distributed through DataLad are organised in. A typical workflow clones a BIDS\ndataset with DataLad, validates it with the BIDS tooling, then runs a BIDS-App under\n`datalad containers-run` so the derivatives carry provenance.\n\n## Primary sources\n\n- DataLad documentation: \u003Chttps:\u002F\u002Fdocs.datalad.org\u002Fen\u002Fstable\u002F>\n- DataLad Handbook: \u003Chttps:\u002F\u002Fhandbook.datalad.org\u002Fen\u002Flatest\u002F>\n- `datalad run` chapter: \u003Chttps:\u002F\u002Fhandbook.datalad.org\u002Fen\u002Flatest\u002Fbasics\u002F101-108-run.html>\n- YODA principles: \u003Chttps:\u002F\u002Fhandbook.datalad.org\u002Fen\u002Flatest\u002Fbasics\u002F101-127-yoda.html>\n- STAMPED principles (operationalized from YODA): \u003Chttps:\u002F\u002Fstamped-principles.org>\n- datalad-container: \u003Chttps:\u002F\u002Fdocs.datalad.org\u002Fprojects\u002Fcontainer\u002Fen\u002Fstable\u002F>\n- git-annex: \u003Chttps:\u002F\u002Fgit-annex.branchable.com\u002F>\n- Dataset registry: \u003Chttps:\u002F\u002Fregistry.datalad.org>\n\n## Acknowledgment\n\nTopic scope for this skill was informed in part by @bcmcpher's MIT-licensed\n[datalad-cli](https:\u002F\u002Fgithub.com\u002Fbcmcpher\u002Fmy-skills\u002Ftree\u002Fmain\u002Fplugins\u002Fdatalad-cli)\nplugin (nineteen per-command slash-command skills). The text here is written\nindependently and grounded in the upstream DataLad documentation; overlap is unavoidable\nbecause both cover DataLad, but the structure, style, and specific technical claims are\ndifferent.\n\n## Agent operating procedure\n\n1. **Check the environment.** Confirm the Python environment and library versions (`python -c \"import pkg; print(pkg.__version__)\"`) and inspect the data's shape, types and missing values.\n2. **Pin down the inputs.** Confirm formats, identifiers and parameters from the data or the user. Ask rather than guess any value that changes the result.\n3. **Run a small version first.** Run on a sample or a single fold first and check runtime and memory.\n4. **Execute the full task** using the instructions and references above.\n5. **Validate the result.** Use held-out data, fixed random seeds and appropriate metrics; check for leakage; report uncertainty (CIs, std over seeds).\n6. **Report.** State what was run (versions, commands, parameters), what was checked, and what is still uncertain.\n\n| If this happens | Do this |\n|---|---|\n| Out-of-memory or very slow execution | Subsample, use chunked or lazy computation, or reduce model size, and tell the user what changed. |\n| A function, flag or endpoint in these instructions is missing in the installed version | Check the installed version's own documentation (`help()`, `--help`, official docs), adapt, and tell the user. Never invent an API. |\n| A required input, identifier or parameter is ambiguous | Ask the user, or state the assumption explicitly before running. |\n\n**Integrity rules**\n\n- Never fabricate results, parameters, identifiers, citations or statistics. If something cannot be run or verified, say so plainly.\n- Never report a metric you did not compute in this session; show the code path that produced every number.\n- Treat version-specific details here as possibly outdated: confirm them against the official documentation for the installed version.\n- Ask before actions that cost money, consume shared GPUs or cloud quota, touch personal or patient data, or cannot be undone.\n",{"name":5,"license":15,"metadata":48,"description":7,"allowed-tools":54,"compatibility":55},{"tags":49,"version":50,"category":51,"maintainer":52,"contributor":53},"research data management, data versioning, provenance, reproducibility, open science","1.0","data-science-and-ml","Kalaris Labs","Dylan Pulver","Read Write Edit Bash","Needs datalad 1.6.x on Python 3.10+, plus git and git-annex 10.x. git-annex is not written in Python but installs as a prebuilt wheel from PyPI (`uv pip install git-annex`), from a system package manager, or from conda-forge. Container-based provenance also needs datalad-container (1.2.x) and Singularity\u002FApptainer or Docker. clone, get, and push need network access; credentialed remotes read secrets from the system keyring or from DATALAD_CREDENTIAL_\u003CNAME>_\u003CCOMPONENT> environment variables.","main","0da4485785ee599388900be50db1635489d53056",".agents\u002Fskills\u002Fkal-datalad","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":66,"prompt":76,"zipAvailable":77},[67,70,73],{"tool":68,"command":69},"npx","npx skills add Tashima-Tarsh\u002FDisha6.6 --skill datalad",{"tool":71,"command":72},"bunx","bunx skills add Tashima-Tarsh\u002FDisha6.6 --skill datalad",{"tool":74,"command":75},"pnpm","pnpm dlx skills add Tashima-Tarsh\u002FDisha6.6 --skill datalad","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fkal-datalad\u002FSKILL.md",true,[79,85,89],{"path":80,"resourceType":81,"resourceTypeDesc":82,"sizeBytes":83,"sourceHtmlUrl":84},"references\u002Fdata-access.md","REFERENCE","参考文档",8492,"https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fkal-datalad\u002Freferences\u002Fdata-access.md",{"path":86,"resourceType":81,"resourceTypeDesc":82,"sizeBytes":87,"sourceHtmlUrl":88},"references\u002Fprovenance.md",11983,"https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fkal-datalad\u002Freferences\u002Fprovenance.md",{"path":90,"resourceType":81,"resourceTypeDesc":82,"sizeBytes":91,"sourceHtmlUrl":92},"references\u002Fpublishing.md",9182,"https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fkal-datalad\u002Freferences\u002Fpublishing.md",[94,126,150,177,201,225],{"skillNo":95,"name":96,"nameZh":6,"description":97,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":98,"categories":99,"keywords":105,"publishedAt":43},"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.","https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fyoutube-full\u002FSKILL.md",[100],{"categoryNo":101,"categoryName":102,"taxonomyType":103,"taxonomyTypeDesc":104},"01JBSEEDCAT000000000000016","图像与多媒体","SCENARIO","使用场景",[106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125],"youtube full","youtube","full","transcript","search","monitoring","data","Video","Audio","list","files","do","get","age","trigger","URL","for","browsing","request","youtube transcripts",{"skillNo":127,"name":128,"nameZh":6,"description":129,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":130,"categories":131,"keywords":132,"publishedAt":43},"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",[],[133,134,135,136,110,137,138,139,140,141,142,143,34,144,145,146,147,148,149,122],"twitter","x twitter growth","growth","research","content","engineering","plan","web search","competitors","analyze","engine","strategy","use","platform","grow","posting","competitive",{"skillNo":151,"name":152,"nameZh":6,"description":153,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":154,"categories":155,"keywords":159,"publishedAt":43},"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",[156],{"categoryNo":157,"categoryName":158,"taxonomyType":103,"taxonomyTypeDesc":104},"01JBSEEDCAT000000000000023","Skill与智能体开发",[160,34,161,162,163,164,165,166,145,167,168,56,169,170,171,172,173,174,175,176],"write a skill","github","create","https","productivity","matt pocock","direct","example","build","new skill","structure","new agent","resources","with","tree","progressive disclosure","author",{"skillNo":178,"name":179,"nameZh":6,"description":180,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":181,"categories":182,"keywords":183,"publishedAt":43},"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",[],[184,185,186,187,188,42,189,190,191,145,192,117,193,194,195,196,197,198,199,200],"wrap up","up","wrap","gemini","deployment","file","hardening","closeout","stop","upgrade","founder","close","companion","you","tables","directions","phase",{"skillNo":202,"name":203,"nameZh":6,"description":204,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":205,"categories":206,"keywords":208,"publishedAt":43},"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",[207],{"categoryNo":157,"categoryName":158,"taxonomyType":103,"taxonomyTypeDesc":104},[209,210,211,212,213,162,42,214,215,216,217,218,34,219,116,220,221,222,223,224],"workflow","workflow builder","builder","design","workflows","code","orchestrate","pipeline","loop","multi-agent","context","across","fresh","scripts","judge","claude code workflow",{"skillNo":226,"name":227,"nameZh":6,"description":228,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":229,"categories":230,"keywords":231,"publishedAt":43},"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",[],[232,233,234,235,236,237,238,39,239,217,240,241,242,243,244,245,118,246,247,200],"review","weekly review","weekly","deterministic scripts","reflect","feeds","inventory","team","projects","project-management","scanner","conversation","back","project","sprint","cadence"]