docs: size Common questions to the evidence, not to a hard observed-only rule

An observed question still beats an invented one and the hunt is still
mandatory, but a thin skill may carry a question a reader would plainly
ask. The count stays honest to the evidence rather than padding every
page out to match a richer one.

.agents/writing-docs.md

@@ -45,13 +45,15 @@ The single non-negotiable: **surface the skill's leading word / defining idea**
 
 The questions readers really ask about this skill, each in bold with the answer in the lines beneath it — no sub-headings.
 
-Every question here is one someone asked. That is the section's whole value: a page with three observed questions beats one with eight plausible ones, because an invented question teaches the reader nothing and reads as filler. So go and find them before you write any:
+An observed question always beats an invented one, so go and find them before you write any:
 
 - **The wiki.** If `~/repos/matt/personal-wiki` exists on this machine, it is the richest source there is. Its `wiki/audience/` area is organised around what the audience wants, discusses, and **is confused by** — read `wiki/index.md` first for the registry of pages, then the pages bearing on this skill. Every page carries `sources:` linkbacks to the original X, Discord, GitHub, and email threads; the wiki is a secondary source, so quote the asker's own question rather than the wiki's summary of it. Skip this bullet where the directory does not exist.
 - **This repo's issues.** `gh issue list --repo mattpocock/skills --search "<skill-name>" --state all`. A question filed twice is a question the page owes an answer to.
 - **`CHANGELOG.md`.** Anything renamed, moved, or behaviourally changed generates a "where did it go?" that the page has to answer.
 
-Order them by how often each comes up, sharpest first, and say the unflattering thing where it is true — a very long grilling session usually means the scope was too big; a model asked to write its own skill produces something verbose. Omit the heading where the hunt turns up nothing.
+Where the hunt comes up thin, the section may also carry a question a reader would plainly ask — but **the count stays honest to the evidence**. A well-discussed skill earns six; an obscure one earns one or two, or none at all. Padding a thin skill out to match a rich one is how the section fills with questions nobody has, and an invented question teaches the reader nothing.
+
+Order them by how often each comes up, sharpest first, and say the unflattering thing where it is true — a very long grilling session usually means the scope was too big; a model asked to write its own skill produces something verbose. Omit the heading where there is nothing worth answering.
 
 ## It's working if
 
@@ -82,7 +84,7 @@ Always present. Situate the skill in the system in a sentence or two:
 - `## Where it fits` names the role and links to `ask-matt`.
 - A prerequisite (workspace, prior setup, tooling) is stated where one exists, and the section is absent where none does.
 - The middle surfaces the leading word.
-- Every `## Common questions` entry traces to a question someone asked — found in the wiki, the issues, or the changelog — and the heading is absent where the hunt found none.
+- The hunt for real questions ran — the wiki, the issues, the changelog — and `## Common questions` is sized to what it found, not padded to match a richer skill's page.
 - Every `## It's working if` bullet is checkable without opening `SKILL.md`.
 - The sections appear in the template's order.
 - Every link is absolute, and every one resolves.