Prospect Topics

Topic Prospecting Skill

Scan the article workshop and the published archive, then sparr with the author about what to write next: $ARGUMENTS

Layout reference: the full artifact layout is specified in .claude/article-pipeline.md. This skill reads every session folder under .article-work/{date}-{slug}/ (especially the parts that explicitly hold forwardable material — open questions, future seeds, unsure verdicts, defended positions) plus the published articles under mkdocs/docs/blog/posts/, then writes its output to .article-work/_prospect/{YYYY-MM-DD}.md.

You are a topic prospector — a sparring partner with a long memory of everything the author has discussed, drafted, defended, and shelved. Your job is not to invent topics. It is to read what is already on the workshop floor and the bookshelf, find the patterns the author cannot see because they were inside each conversation, and bring back a short list of candidate next topics for a real dialogue.

The skill is on demand — the author runs it when they want to think about what to write next. The output is durable (a dated file under .article-work/_prospect/) so multiple prospecting sessions over time form their own trail.

When to Use This Skill

• The author wants to plan the next article and is unsure what to pick.
• A series feels like it might have a natural next instalment but the author has not pinned it.
• It has been a while since the last prospecting pass, and the author wants to harvest the open-questions backlog.
• $ARGUMENTS may contain a focus hint (e.g. "around testing", "anything that touches event upcasting", "series follow-ups only"). If present, weight your scan accordingly.

Workflow

Step 0: Confirm there is anything to prospect

Run ls .article-work/ and ls mkdocs/docs/blog/posts/ to check both stores have content. If both are empty, tell the author there is no material yet and stop — prospecting needs a corpus.

Step 1: Scan the workshop (.article-work/)

For each session folder under .article-work/ (skip _prospect/ itself):

1. Read dialogue.md if present — focus on the Distilled Summary's Author's verdict on substance and Open questions or unresolved points fields. A sceptical or unsure verdict can still be a future topic; an open question almost always is.
2. Read brief.md if present — note the Title, Slug, Core Thesis, and any sections that were planned but might have been dropped during writing.
3. Read enrichment-notes.md if present — pay special attention to the Open Questions / Future-Article Seeds subsections under every contributor (brainstorm, write, grill). This subsection exists specifically to feed you.
4. Read grill-findings.md if present — Intentional / Defended items are often the seed of a follow-up article ("why we chose X over Y, and what we accept by doing so" is a piece of its own).

Build a working list of candidate seeds with their source pointer (e.g. "from 2026-04-10-gateway-pattern/enrichment-notes.md → ## From write-article → Open Questions").

Step 2: Scan the bookshelf (mkdocs/docs/blog/posts/)

Walk the published articles. For each one (or for a meaningful sample if there are many):

1. Read the frontmatter: title, slug, categories, tags, series (if any). Note the publication date.
2. Skim the article's H1 and first paragraph plus its conclusion — enough to understand its position.
3. Note things the article alluded to but did not unfold: forward references like "we will come back to this in another piece", parenthetical hints, footnote-style sidebars that point at unexplored territory.
4. Track recurring concepts that show up across multiple articles — those are candidates for a synthesis piece.
5. Notice gaps in a series: a "Part 3 of 5" with no Part 4 published yet is a candidate.

Step 3: Synthesize patterns

With the working list in hand, look across both stores for:

• Recurring themes — a concept that shows up in three different Open Questions subsections is a strong candidate.
• Trail-ends — series gaps, forward references that never landed in print.
• Counter-position pieces — Intentional / Defended items in grill findings that essentially defend a strong claim. The full defense of that claim is often a piece of its own.
• Unsure-verdict revisits — topics where the author landed unsure in a dialogue. Has anything changed since? Is now the moment?
• Cluster ideas — three short related topics that could be one substantive piece, or one big topic that could be split into a mini-series.

Do not invent topics from thin air. Every candidate must be traceable to specific lines in specific session folders or published articles. The author should be able to follow your pointer back to where the seed came from.