Add article planning interview skill.

Author: Jacques de Vos

.agents/skills/plan-article/SKILL.md

@@ -0,0 +1,139 @@
+---
+name: plan-article
+description: >-
+  Interviews the author to uncover and develop an article idea, researches
+  relevant background, tests the idea for gaps, and builds a rich article plan
+  using the site's structure specification. Use when the user wants to explore,
+  interrogate, or plan a blog article before drafting it.
+disable-model-invocation: true
+---
+
+# Plan Article
+
+Develop the author's thinking before writing prose. The result is a rich, reviewable article plan, not a draft.
+
+## Start
+
+1. Switch to Plan mode and remain read-only.
+2. Read `structure-spec.md`.
+3. Treat its spine as a coverage guide, not a questionnaire or rigid outline.
+4. Begin with one broad invitation, such as: “Tell me about the idea and the problem you think it solves.”
+
+Ask one main question at a time. Prefer open questions that let the author speak at length. Follow interesting threads before returning to gaps. Do not turn the conversation into a form, rapid-fire checklist, or premature outline.
+
+## Conversation style
+
+Make the interview feel like a thoughtful conversation with an interested peer:
+
+- Keep each turn short enough that the author remains the main speaker.
+- Respond to what the author just said before asking the next question.
+- Base follow-up questions on the author's words, examples, uncertainties, or contradictions rather than reciting this skill's categories.
+- Use the author's terminology unless a distinction needs clarification.
+- Ask simple questions with one clear centre of gravity. A brief prompt such as “Can you give me an example?” is often enough.
+- Let useful tangents develop. Quietly track uncovered areas and return to them when the current thread slows down.
+- Do not announce phases, explain the interview framework, or list all the questions still to come.
+- Do not use a structured multiple-choice question for exploratory discussion. Reserve it for a genuine decision that cannot be inferred.
+- Avoid generic praise, repeated paraphrasing, and an interviewer's commentary after every answer.
+- Challenge assumptions and factual claims respectfully when doing so would sharpen the idea.
+
+The first response should contain only a brief welcome and the opening invitation. Do not front-load instructions or an agenda.
+
+## Interview
+
+Move through these phases naturally. Revisit earlier phases whenever the author's thinking changes.
+
+### 1. Open exploration
+
+Understand, in the author's own terms:
+
+- the idea and why it is interesting now
+- the problem, tension, or opportunity behind it
+- the experiences or observations that prompted it
+- what the author suspects but has not resolved
+
+Do not narrow the idea too quickly. Reflect useful phrases and ask short follow-ups that draw out examples, stories, distinctions, and consequences.
+
+### 2. Context and reader
+
+Once the initial thinking is visible, explore:
+
+- the specific engineering reader and what they already know
+- what that reader wants to understand, reconsider, decide, or try
+- relevant context, constraints, history, and vocabulary
+- who experiences the problem, when it appears, and why current approaches fall short
+- the stakes if the problem is ignored and the opportunity if it is addressed
+
+Ask for concrete incidents and counterexamples. Separate what the author experienced from what they heard or inferred.
+
+### 3. Idea or proposed approach
+
+Invite the author to explain the solution freely, then probe:
+
+- the core model, mechanism, or decision rule
+- how it differs from familiar approaches
+- examples that make it concrete
+- where it works, where it does not, and what it requires
+- likely objections, failure modes, unintended consequences, and tradeoffs
+- credible alternatives and the strongest case for each
+
+Point out meaningful gaps and tensions directly, but do not force certainty where the author is still exploring.
+
+### 4. Evidence and wider perspectives
+
+Ask what supports the idea: direct experience, examples, data, sources, other people's arguments, or reasoned inference. Clearly label:
+
+- experience
+- sourced evidence
+- inference
+- opinion
+- open question
+
+When the author mentions where an idea came from, a source, person, paper, product, or related debate, offer to research it on the web. Use research to deepen or challenge the thinking, not to replace the author's perspective. Record links and distinguish source claims from the author's claims. Never invent evidence, experience, certainty, or opinions.
+
+### 5. Understanding checkpoints
+
+Occasionally, after several useful exchanges or when the idea materially changes, give a short bullet summary beginning with a question such as:
+
+“Is this what you are trying to convey?”
+
+Summarise:
+
+- the central idea as currently understood
+- the problem and why it matters
+- the intended reader and desired effect
+- the strongest support
+- unresolved tensions or gaps
+
+Then invite corrections, missing perspectives, and further thoughts. Treat the author as authoritative about their intended meaning, while continuing to verify external facts and test the reasoning.
+
+## Use the article spine
+
+Privately check coverage against:
+
+1. audience, reader goal, and intended outcome
+2. central idea
+3. problem, opportunity, and necessary context
+4. proposed approach and concrete examples
+5. alternatives, evidence, and tradeoffs
+6. assumptions, limitations, and open questions
+7. a possible transferable takeaway or decision rule
+8. visual concepts that could strengthen the idea
+
+Do not spend interview time finalising front matter, title, TL;DR, conclusion, image, or polished section wording. Capture promising options, but drafting happens later.
+
+## Final article plan
+
+When the author agrees that enough material has been gathered, produce a detailed plan containing:
+
+- **Working idea:** a concise statement of what the article is trying to convey
+- **Reader and outcome:** who it is for and what should change for them
+- **Problem and opportunity:** context, stakes, constraints, and concrete examples
+- **Proposed approach:** model, mechanism, examples, and practical implications
+- **Evidence map:** experience, sourced evidence with links, inference, and opinion
+- **Alternatives and tradeoffs:** credible competing views, objections, limits, and failure modes
+- **Open questions:** uncertainties, missing evidence, and ideas worth exploring
+- **Possible article spine:** a flexible sequence of sections with the purpose of each
+- **Useful material:** stories, metaphors, quotes, terms, examples, and possible visuals
+- **Writing compass:** voice notes, claims requiring verification, and ideas that belong in separate posts
+
+Keep the plan expansive enough to preserve valuable thinking. Do not turn it into article prose. End by asking the author to review, correct, and add to the plan. Only hand off to a drafting skill when the author explicitly says the plan is ready.