marketing: FRE-199 first pair of informational pages + pitch shape (drafts)

- agent-done-signal: 'How do you know when your coding agent is done?' —
  the zero-looks-like-success cluster (1xjb3j0, 1xm09fs, 1xmnz2c, HN fhub/
  sadilek); Stop/Notification/StopFailure hooks cited to the agent's own
  docs, re-read 2026-11-05. No product, no competitor, no price.
- agent-waiting-on-permission: 'Why is my coding agent sitting on a
  permission prompt?' — the 1xhbc71 thread + invisible-permission cluster;
  permission modes, /permissions, rule syntax, acceptEdits cited to the
  permissions doc; undo-split attributed to 1xkoz06.
- newsletter-postmortem-pitch: item 2's shape, placeholders only; send
  gated on FRE-202 or a CTO-confirmed direct relay-store read.

All drafts: sponsor fact-check moves pending/ -> approved/; nothing
publishes itself.

marketing/drafts/newsletter-postmortem-pitch/copy.md

@@ -0,0 +1,82 @@
+# Newsletter pitch — post-mortem shape (placeholders; numbers gated)
+
+> **Status: draft — placeholders only. No figure in this file is a measurement.**
+> FRE-199 item 2. Drafted 2026-11-05 by CMO; the plan's date was Fri 2026-11-06.
+> **Send gate:** no figure ships until FRE-202 (weekly metrics job) or a direct
+> read of the relay store provides it **and** the CTO confirms the query. Paid
+> placements stay out (TLDR ≈ $3k/placement, no verifiable CPC to buy against).
+
+## Audience & channel
+
+Editors of free dev-tool newsletters (the ones that run post-mortems and
+"what we learned" pieces). Direct email, one editor at a time, personalised.
+No paid placements at any budget without a CoS + sponsor decision.
+
+## The claim the pitch goes out with
+
+A post-mortem pitch, not a product pitch: *here is what shipping a Mac utility
+alone actually broke, and what we changed because of it.* The product is the
+setting; the breakage and the fix are the story.
+
+## The pitch (shape — every figure is a placeholder)
+
+**Subject:** What shipping a Mac utility alone broke (post-mortem)
+
+Hi <editor>,
+
+We build The Nook, a $10 Mac utility that surfaces coding-agent approvals at
+the notch. I'm not pitching the app — I'm offering a post-mortem your readers
+usually only get from funded startups, because shipping alone broke things in
+ways bigger teams don't hit:
+
+- **The silent update.** In October we shipped a release whose appcast published
+  cleanly — while every delta it named 404'd, so users on older builds were
+  being offered updates that didn't exist. The publish path deleted files
+  before the feed that named them was live. Nothing alerted; the feed just
+  lied quietly. (Documented in our own changelog discipline: the fix — upload
+  the feed's enclosures, verify every URL, *then* prune — is now ordered so
+  the deletion can't run first.)
+- **[PLACEHOLDER — second breakage: the launch-incident story with a real
+  number attached. Candidates from the engineering journal: the 0.13.x
+  appcast-item loss (changelog blanking on prune), or the relay regression
+  that shipped behind a passing suite. Pick the one FRE-202's data can
+  quantify.]**
+- **The numbers, as they are.** <N> activations and <X>% trial→paid as of
+  <date> — small numbers, real ones, and the interesting part is the funnel
+  shape, not the size.
+
+The angle for your readers: solo distribution is a chain of quiet failure
+modes, each one invisible until it costs a user. We kept the receipts.
+
+Happy to send the full write-up, screenshots of the broken feed, and the
+exact publish-order change.
+
+— <name>, FrenoCorp
+
+## Where the numbers come from (and the caveats that travel with them)
+
+- **Preferred source:** FRE-202, the weekly metrics job (CTO owns the query).
+- **Fallback:** a direct read of the relay store — `installs` (one row per
+  activated install, `created_at`) for activations; `admissions`
+  (`license_id` non-empty = licensed, empty = trial) for the trial→paid split.
+- **Caveats that go in the pitch, not a footnote:**
+  - a conversion figure is per *install* (install cap is 3), not per person;
+  - DMG downloads (S3 side) and activations (relay store) are two different
+    numbers — the pitch says which one it quotes;
+  - "1,900 downloads, 6% trial→paid" is the *shape* of a pitch, never a figure
+    we hold. No number is estimated to fill a placeholder.
+
+## Gates before anything is sent
+
+1. Figures exist (FRE-202 or CTO-confirmed direct read) — owner: CTO.
+2. CTO confirms the query behind every figure quoted — owner: CTO.
+3. Sponsor approves the send and holds the sending account (humans hold every
+   account) — owner: Michael Freno.
+4. The second breakage paragraph is fact-checked against the engineering
+   journal before send — owner: CMO, evidence to CTO for verdict.
+
+## Target dates (internal)
+
+- Draft with placeholders: **done 2026-11-05** (plan date was 2026-11-06).
+- Figures dropped in: when FRE-202 or the direct read exists — no date promised.
+- Send: only after gates 1–4. Sponsor submits.

marketing/drafts/newsletter-postmortem-pitch/rationale.md

@@ -0,0 +1,49 @@
+# Rationale — newsletter post-mortem pitch (shape)
+
+FRE-199 item 2. Drafted 2026-11-05 by CMO (plan date was 2026-11-06; FRE-201
+landed today, which was the only gate on drafting the shape).
+
+## Why draft now
+
+The Chief of Staff's sequencing had the pitch drafted when FRE-201 landed. It
+landed 2026-11-05 (row closed `done` 18:45Z; the daily scan routine
+`2553b201` has been producing the backlog since 2026-10-30). The figures were
+never gated on FRE-201 — only the shape was — so the shape ships now and the
+figures stay placeholders behind their own gate.
+
+## Why a post-mortem and not a product pitch
+
+The issue text's own thesis: "1,900 downloads, 6% trial→paid, here's what
+broke" outperforms a product pitch. Editors of free dev-tool newsletters run
+post-mortems; a solo shop shipping a Mac utility and keeping receipts is a
+story they can run without us asking for a review.
+
+## The breakage is real, not invented
+
+The lead paragraph is the 0.13.4 delta incident (FRE-542): the appcast stayed
+at 0.13.4 while its five delta files 404'd — the publish path pruned before
+the feed was up. The fix (upload enclosures, verify URLs, copy the appcast,
+then prune) is in `scripts/upload-release.sh` and is engineering-verified by
+the FRE-542 record. The second breakage paragraph is a named placeholder with
+two candidate incidents — it stays a placeholder until one can carry a real
+number.
+
+## Hard rules honoured
+
+- Every figure is `[PLACEHOLDER]` and labelled as such; the file says in its
+  header that no figure in it is a measurement.
+- No invented number, no estimated conversion, no user counts.
+- The two caveats (per-install conversion with install cap 3; downloads ≠
+  activations) are written to travel inside the pitch, not a footnote.
+- Paid placements: out, per the issue text and the plan (TLDR ≈ $3k, no
+  verifiable CPC). No placement is bought without CoS + sponsor.
+- Price appears once ($10, what the app actually charges) to identify the
+  product — no discount, no countdown, no price-change copy (FRE-204 holds).
+
+## Gates before send
+
+1. Figures from FRE-202 or a CTO-confirmed direct relay-store read — CTO.
+2. CTO confirms the query behind every figure — CTO.
+3. Sponsor approves the send; a human holds the account and sends — sponsor.
+4. Second-breakage paragraph fact-checked against the journal — CMO drafts,
+   CTO verifies.