Skill v1.0.0

version: "1.0.0" name: safely-unpack-zip description: Unpack zip files inside a Docker container (via Colima) and serve over HTTP. Nothing touches the local filesystem. argument-hint: "[directory]" allowed-tools:

• exec
• read
• grep
• glob

Unpack all zip files in $ARGUMENTS (default: ~/Downloads) and serve them over http://127.0.0.1:8080 — entirely inside a Docker container. Nothing is extracted locally.

Colima resource notes

Colima runs Docker in a Linux VM on macOS. The VM has a sparse virtual disk (grows on demand, does not reserve full size on host) and fixed RAM/CPU allocation.

Current profile defaults (verify with colima list):

• Disk: 100GiB ceiling (sparse — only grows as containers write data)
• Memory: 8GiB RAM
• CPUs: 2

Practical limits:

• Extracted zip contents must fit within free VM disk space. Check with docker system df if unsure.
• unzip is light on RAM — won't hit 8GiB ceiling unless zips are very large.
• No per-zip size guard in this workflow. If extraction fails silently, suspect VM disk exhaustion.

To reclaim VM disk space (removes unused images, stopped containers, dangling volumes):

docker system prune -a

"Tear it down"

When Fahad says "tear it down", stop everything and return all storage:

1. Kill the serving container: docker ps -q --filter "publish=8080" | xargs docker kill
2. Prune all unused Docker data: docker system prune -a -f
3. Stop Colima: colima stop

This frees VM disk back to baseline (only images/layers you keep remain). Colima's sparse disk shrinks on the host as the VM releases space.

Steps

1. Ensure Colima is running. If not, start it:

`` colima start ``

1. Check ~/.docker/config.json — if "credsStore": "desktop" is present, remove it (breaks Docker without Docker Desktop).

1. Find all .zip files (type f only, not directories) in the target directory:

`` find <dir> -maxdepth 1 -name "*.zip" -type f ``

1. Kill any container already holding port 8080:

`` docker ps -q --filter "publish=8080" | xargs docker kill ``

1. Build a docker run command that:

• Mounts each zip as -v "<abs_path>:/tmp/zips/<name>.zip:ro"
• Uses --cap-drop ALL and -p 127.0.0.1:8080:8080
• Uses python:alpine image
• Runs: for z in /tmp/zips/*.zip; do name=$(basename "$z" .zip); mkdir -p "/tmp/out/$name"; unzip -o "$z" -d "/tmp/out/$name"; done && cd /tmp/out && python -m http.server 8080

1. Run the container in the background. Verify with:

`` curl -s http://127.0.0.1:8080/ `` Confirm each chapter directory appears in the listing.

Notes

• Always use absolute paths in -v mounts — never $(pwd) inside exec tool calls.
• Quote all paths with spaces.
• Each zip unpacks into its own subdirectory named after the zip (minus .zip).
• Server is localhost-only (127.0.0.1:8080), not exposed to the network.
• Before running, group related files into zips (e.g., Chapter 1.pdf + Chapter 2.pdf → Chapters.zip). Delete source files after zipping.
• If extraction fails, check docker system df — VM disk may be full. Run docker system prune -a to reclaim.