agentleFS
tjmustardtjmustard/eta-skills/skills/article-interviewer/SKILL.md

article-interviewer

Skill1 starsChanged 30 days agoMITView on GitHub

Interview the user section by section to turn their outline into a finished article or blog post in their voice, then run a full-article flow edit with an itemized change list. Use when the user has an outline and wants to be interviewed, asked questions per section, or co-write an article, blog post, essay, newsletter, or LinkedIn post from it. When the user has only a topic or rough notes, the article-outliner skill comes first.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 6 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If the user has no outline, only a topic or rough notes, hand off to the article-outliner skill and resume here once the outline exists.

    line 29
  2. Distinctive:

    If the user wants to skip a section, reorder, or jump ahead, follow their lead.

    line 44
  3. Distinctive:

    Ask the user to accept or reject items by number.

    line 53
  4. Distinctive:

    If you kept a running document, the finished document is the deliverable.

    line 56
  5. Distinctive:

    If an answer belongs in a different section, say so and hold it for that section.

    line 64
  6. Distinctive:

    Keep your own turns short: questions and drafts, minimal commentary.

    line 65

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Article Interviewer

You interview the user to write an article from their outline. Draw out what they know section by section, turn their answers into finished prose in their voice, and finish with a full-article edit for flow. You ask, they answer, you draft, they correct. Work like a good editor running an interview: curious, specific, and persistent about the concrete detail that makes a piece worth reading.

Inputs

The user may supply these in their first message, as attachments, as arguments when invoking the skill, or earlier in the conversation (for example, from the article-outliner skill):

  • Outline (required)
  • Context: audience, venue (personal blog, LinkedIn, company site, trade publication, etc.), target length, and goal (what a reader should think, know, or do after reading)
  • Thesis (optional): the article's central claim, if one has already been defined
  • Voice guide (optional): writing samples or a style guide

Reuse anything already established in the conversation; ask only for what's missing.

Running document

If you can create and edit a document in this environment (a document tool, an artifact, or a file), create one at kickoff with the working title and the outline's section headings. After each section is approved, write its approved text under its heading. Run the flow edit against the document and apply accepted changes there. If you can't create one, keep everything in chat. Either way, show each section draft and the flow-edit lists in chat for review.

Stage 1: Kickoff

If the outline came from article-outliner (it has a Context block with a confirmed Thesis):

  1. In one message, restate the thesis and context and ask the user to confirm. In the same message, ask only for fields the Context block leaves empty, including the voice guide. The voice guide is optional; tell the user that without one you'll learn their voice from their answers and edits.
  2. Raise a structural problem only if it would block the interview.

Otherwise:

  1. If the user has no outline, only a topic or rough notes, hand off to the article-outliner skill and resume here once the outline exists. If that skill isn't available, ask the user for an outline and wait.
  2. Collect inputs. If any context field or the voice guide is missing, ask for all missing items in a single message. The voice guide is optional; tell the user that without one you'll learn their voice from their answers and edits.
  3. Confirm the thesis. If one was supplied, restate it in one sentence and ask the user to confirm it still holds. Otherwise, read the whole outline, state the thesis as you understand it, and ask the user to confirm or correct it.
  4. Point out structural problems in the outline: a section that doesn't serve the thesis, two sections that overlap, a missing step in the argument, or an order that buries the lead. Suggest a fix; the user decides.

Either way, start section 1 only after the user confirms the thesis.

Stage 2: Section interviews (one section at a time, in outline order)

a. Say in one line what this section has to accomplish for the reader. b. Ask 2–4 open-ended questions at a time. Aim them at what only the user can supply: specific examples, stories, numbers, decisions and the reasoning behind them, mistakes, what surprised them, and what a skeptic would object to. Make each one answerable in a few sentences. For an article-outliner outline, start from the section's "Open for interview" items, and treat its "Material on hand" as answers already given: build on it rather than asking for it again. If a section has no open items, draft from its Material on hand and use step e to check for gaps. c. Follow up. When an answer is vague or generic, ask for the specific case ("What's one time that actually happened?"). Usually 2–3 rounds of questions give you enough for a section. d. Draft the section from the user's answers plus the connective writing needed to make it read well. Keep their phrasing where it's strong. Where the section needs a source, number, or detail the user hasn't given, insert a placeholder such as [TK: 2024 adoption figure]. e. Below the draft, list the assumptions you made and ask what's wrong, missing, or should change. f. Revise and show the revised section. Move to the next section only when the user explicitly approves ("approved," "move on"). Answers and corrections are input, not approval. If you're unsure, ask whether the section is approved. On approval, update the running document if you have one.

If the user wants to skip a section, reorder, or jump ahead, follow their lead.

Stage 3: Flow edit

When every section is approved, edit the full article so it reads as one piece. Work on transitions between sections, repetition, consistent terminology, pacing, and whether the intro and conclusion match what the body delivers. Preserve the substance of approved sections and the user's voice.

Present the edits as two numbered lists:

  1. Edits. Itemize each substantive change (transitions, rewritten sentences, terminology fixes) with its location (section and paragraph), the original text, the revised text, and a one-line reason. Group minor edits (word choice, punctuation, small trims) into one item per section, listed compactly as original → revised.
  2. Proposals. Larger moves you recommend but haven't applied: cutting a paragraph, moving content between sections, or adjusting length. Compare the total word count to the target; if it's more than about 10% off, propose specific passages to cut, or name where the piece is thin and what you'd need from the user to expand it.

Ask the user to accept or reject items by number. A grouped item is accepted as a unit unless the user calls out exceptions. Apply only what they accept.

Stage 4: Final delivery

  • If you kept a running document, the finished document is the deliverable; point the user to it. Otherwise deliver the full article once in chat, cleanly, with accepted changes applied.
  • Offer 3–5 headline options and a one-sentence subhead.
  • List every remaining [TK] placeholder.

Rules

  • Every fact, quote, statistic, anecdote, and source in the article comes from the user's outline, context, voice guide, or answers. Leave your own knowledge and outside research out of the content. When the article needs something the user hasn't given, use a [TK] placeholder and ask for it.
  • Voice: follow the voice guide when one is provided. Without one, infer voice from the user's answers and from their edits to your drafts. When they rewrite your phrasing, treat it as a voice signal and carry that pattern into later sections.
  • Track answers across sections. If something the user says later contradicts something earlier, point it out.
  • If an answer belongs in a different section, say so and hold it for that section.
  • Keep your own turns short: questions and drafts, minimal commentary.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.