| name | ai-at-work-entry |
| description | Add or update entries in the AI at Work section of this blog. Use when the user wants to curate an external tweet/X post, blog post, interview, talk, podcast, or similar source into `_ai_at_work/`, either from a URL or from manually provided metadata. |
AI at Work Entry
Create content for the repo's AI at Work collection in _ai_at_work/.
Use this skill when
- The user wants to add an
AI at Work item from a URL.
- The user wants to add an
AI at Work item from notes or partial metadata.
- The user wants to turn an approved GitHub issue submission into an
_ai_at_work/ entry.
- The user wants to refresh metadata, tags, or the TL;DR/commentary for an existing
_ai_at_work/ entry.
Output target
- Collection directory:
_ai_at_work/
- One file per entry, named from the final slug:
_ai_at_work/<slug>.md
- Prefer the repo scaffold helper if it exists:
just new-ai-at-work-entry "Title"
- If the helper does not exist, create the file directly with the same front matter shape.
Required front matter
Always set:
title
slug
publishedAt
type
sourceUrl
author
externalPublishedAt if available, otherwise leave it blank
commentary
tags
Support these when useful:
sourceName
authorUrl
summary
quote
image
draft
The collection has a default layout in _config.yml, so you usually do not need
to set layout explicitly.
Workflow
If the user provides a URL
- Read the source on the web.
- Extract the title, source name, author, source type, original publish date, and a short summary if it is clearly available.
- Choose 2-5 focused tags. Prefer tags already used by the section, such as
ai-adoption, engineering, agents, workflows, leadership, org-design, ai-coding, productivity, tooling, culture, metrics.
- Default
commentary to a neutral TL;DR, not a first-person personal note.
Only write it in the blog author's voice if the user explicitly provides or asks for that.
- Create the entry file and keep the body optional. If there is a strong excerpt, put it in
quote; do not paste large chunks of the source.
If the user provides a GitHub issue or issue body
- Treat the issue as intake context, not publish-ready copy.
- Use the submitted URL as the source of truth for title, source metadata, and dates when possible.
- Preserve any submitter-provided TL;DR or relevance notes only as raw input to refine; do not present them as the blog author's personal note unless explicitly approved.
- Keep the final
commentary field neutral by default.
- Reuse suggested tags when they fit the section's existing conventions.
If the user does not provide a URL
- Derive a slug from the title.
- Use the provided metadata.
- Leave unknown optional fields blank rather than inventing them.
- If the user asks for a TL;DR, keep it neutral and concise.
- If the user wants a personal note, use only wording the user provided or approved.
Writing rules
- Use
commentary as a short TL;DR by default. Usually 1-3 sentences.
- Do not invent personal opinions, reactions, or first-person framing.
- Do not treat issue submitter text as the maintainer's voice unless the maintainer explicitly says to publish it that way.
- Prefer paraphrase over long quotation.
- Keep tags source-aware and sparse.
- Use
type values that match the section conventions: tweet, blog-post, podcast, interview, talk, paper, other.
- Do not build import automation, scraping pipelines, or extra taxonomy pages as part of adding an entry.
Verification
- If you changed templates, front matter shape, or rendering-related fields, run the site build after editing.
- If you only added a content file and the existing schema already supports it, a full build is still preferred before wrapping up.