Docker Container Rerun

Update a docker run container with a conservative workflow.

Required Inputs

Require both of these from the user:

• - INLINECODE1
• INLINECODE2

Treat recreate_command as the source of truth. Do not try to reconstruct missing flags from docker inspect.

Scope

Support only containers originally managed by docker run.

Do not use this skill for:

• - INLINECODE6
• guessing or synthesizing missing run flags
• deleting volumes
• INLINECODE7
• changing environment variables, mounts, ports, labels, or networks unless the user explicitly changed the recreate command

Update Rule

Always compare image Id values, not repo digests.

Use this exact logic:

1. 1. Read current image Id from the running container:

   docker inspect -f '{{.Image}}' <container_name>
   

1. 2. Extract the image reference from recreate_command.
2. Pull the latest version of that image:

   docker pull <image>
   

1. 4. Read the latest local image Id:

   docker image inspect <image> --format '{{.Id}}'
   

1. 5. Recreate the container only if the two Id values differ.

Safety Rules

Before any destructive action, restate the exact recreate command that will be used.

If recreate_command is missing, ambiguous, or not clearly a docker run command, stop and ask the user to provide a valid full command.

If the image cannot be extracted from recreate_command, stop and ask the user to provide the image explicitly inside the command.

Never silently modify the recreate command.

Prefer this sequence when update is needed:

CODEBLOCK3

Validation of recreate_command

Before using it, verify all of the following:

• - starts with INLINECODE12
• includes an image name as the final image argument before any container command
• clearly targets the same logical container the user wants updated

If the command includes an inline container command after the image, preserve it exactly.

If the command is multiline, preserve it exactly.

Recommended Execution Workflow

1. 1. Confirm the target container name.
2. Echo back the exact recreate command.
3. Extract the image from the recreate command.
4. Compare current image Id and latest pulled image Id.
5. If Ids match, report that the container is already up to date and do nothing else.
6. If Ids differ:

1. 7. Verify startup with:

   docker ps --filter name=<container_name>
   docker inspect <container_name>
   docker logs --tail 100 <container_name>
   

1. 8. Report status clearly, including whether healthcheck is healthy, starting, or absent.

Bundled Script

Use the bundled script when you want a deterministic check/apply flow:

CODEBLOCK5

Add --apply only when the user has approved the exact recreate command and actual recreation should happen.

The script will:

• - validate INLINECODE19
• extract the image
• pull the latest image
• compare current vs latest image Id
• optionally stop/remove/recreate
• emit JSON summary with container state, health status, and recent logs

Output Expectations

When reporting results, include:

• - target container name
• extracted image name
• current image Id
• latest image Id
• whether recreation was needed
• post-recreate container state
• health status if present
• any obvious log errors seen in recent logs

Example Pattern

Input:

• - container_name: INLINECODE21
• INLINECODE22:

Expected behavior:

• - extract image INLINECODE23
• compare current container image Id vs pulled latest image Id
• recreate only if the Ids differ
• preserve the recreate command exactly

Notes

When users ask to "update container X", prefer asking for the original docker run command unless it is already documented in memory or provided in the current request.

If the user has a known fixed recreate command for a specific container, prefer using that exact command unchanged.