workflow-image-readme-sync

name	workflow-image-readme-sync
description	Scan .github/workflows for Docker images built and pushed by this repository, then update the README.md image table with the discovered images, base images, build contexts, workflow paths, and concise agent-written descriptions based on each build context. Use this when asked to sync repository Docker image documentation from GitHub Actions workflows. Use when this capability is needed.
metadata	{"author":"evindunn"}

Workflow Image README Sync

Use this skill when the user wants README.md updated from Docker image definitions in .github/workflows.

Workflow

1. Run python3 scripts/update_readme.py --print-discovery-json to discover images, base images, build contexts, and workflow paths.
2. Read each discovered build context, starting with its Dockerfile and only the nearby files needed to understand the image's purpose.
3. Write concise human descriptions in the style of product summaries, not low-level Dockerfile narration.
4. Save the descriptions in a JSON mapping from image name to description, then run python3 scripts/update_readme.py --description-file <file>.
5. Check the updated table in README.md and verify each row includes Image, Base Image, Context, a Workflow badge that links to the GitHub Actions workflow, and Description.

Notes

• The updater rewrites only the content between the <!-- docker-images-table:start --> and <!-- docker-images-table:end --> markers in README.md.
• The Python script is intentionally data-oriented. It discovers image metadata and renders the table, but the agent is responsible for the final description quality.
• The repository-level scripts for this workflow live in ../../../scripts. Prefer ../../../scripts/update_readme.py and ../../../scripts/generate_readme_image_descriptions.py.
• The Workflow column should render GitHub Actions badge images that link to each workflow page, rather than plain text workflow paths.
• Prefer short summaries like Debian base image with custom ca certs or Vault running in agent mode with config templating.
• If no images are found, the table is populated with a single None row so the README stays explicit.

---

Source: evindunn/docker — distributed by TomeVault.