Datalad Container Run

Skill: datalad-container-run

Wrap a command in datalad containers-run to record the container image, the command, inputs, and outputs in the dataset history. The container image content hash is annexed, so the full computational environment is reproducible via datalad rerun.

Always load ${CLAUDE_PLUGIN_ROOT}/references/container-run.md before any container registration or container-run step.

Steps

1. Identify command and container — read $ARGUMENTS or extract from conversation

context. Determine:

• The container image (.sif path, Docker image name, Singularity/Apptainer URL)
• The command to run inside the container

If either is ambiguous or missing, ask the user to specify both before continuing.

1. Verify DataLad context — check whether cwd is inside a DataLad dataset:

``bash ls .datalad/ ``

• Dataset found: continue.
• No dataset found: ask the user:

> "No DataLad dataset detected. Would you like to: > 1. Initialize one here with /datalad-init > 2. Run the container command bare (no provenance tracking)" Wait for their choice.

1. Check for unsaved changes — run datalad status. If the output shows modified

or untracked files, prompt the user to save first with /datalad-save or confirm they want to proceed anyway.

1. Container registration check — run datalad containers-list to see registered

containers:

• Container already registered: proceed to step 5.
• Not registered: guide the user through datalad containers-add:

a. Ask for a short name (identifier) for the container. b. Confirm the image URL or local .sif path (--url). c. Ask which runtime to use; load ${CLAUDE_PLUGIN_ROOT}/references/container-run.md and show the call-format options table (Singularity, Apptainer, Docker). d. Construct and display the datalad containers-add command: `` datalad containers-add \ --url \ --call-fmt "" `` e. Execute after user confirmation. On failure, show error and stop.

1. Gather run parameters — ask the user (or infer from context) for:

• -i (inputs): files or glob patterns the command reads (omit if none)
• -o (outputs): files or glob patterns the command writes (required if outputs exist)
• -m (message): a meaningful description of what this run does (required)

1. Construct and show the full command — build the complete invocation and display

it before executing: `` datalad containers-run \ --container-name \ -m "" \ [-i ...] \ [-o ...] \ "" `` Then ask: > "Ready to execute? (yes / edit)"

1. Execute and report — after user confirmation, run the constructed command.

• On success: report the commit hash (includes container image hash), output files

recorded, and that the container image and command are now in dataset history.

• On failure: show error output; note that nothing was committed.

Removing a registered container (datalad containers-remove)

When the user wants to deregister a container:

1. List registered containers to confirm the name:

``bash datalad containers-list ``

1. Warn the user — removing a container drops its registration and annexed image

from the dataset. Runs that used this container remain in history but cannot be replayed (the image would need to be re-added): > "Removing '' will drop the container image from the dataset. Past runs > referencing it will no longer be replayable unless you re-add the image. Proceed?"

1. Remove after confirmation:

``bash datalad containers-remove ``

1. Save the change:

``bash datalad save -m "remove container " ``

Constraints

• Always load ${CLAUDE_PLUGIN_ROOT}/references/container-run.md before any container

registration or container-run step.

• Never call datalad containers-run with an unregistered container name — always verify

with datalad containers-list first.

• Never skip showing the full constructed command before executing.
• Always require a meaningful -m message — never use empty or placeholder messages.
• For pipelines inside the container, wrap as "bash -c 'cmd1 | cmd2'" passed to the

container call.

• The --call-fmt must contain both {img} and {cmd} placeholders — if either is

missing, the container will be invoked incorrectly. Flag this to the user.

• For containers-remove: always confirm before executing and always follow with

datalad save to record the removal.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: neuroanalytics
• Source: neuroanalytics/data-science-harness
• License: MIT

Install and usage instructions live in the source repository linked above.