chore(distribution): standardise tool count on 19 + amend cascade docs (Glama auto-publish verified)

CHANGELOG.md

@@ -5,6 +5,7 @@ Subscribe via RSS: <https://www.freightutils.com/changelog.xml>
 
 ## 2026-05-17
 
+- **Distribution**: Tool-count standardised on **19** across all in-repo copy + cascade docs amended. `SITE_STATS.toolCount` and `SITE_STATS.mcpToolCount` in [`lib/constants/siteStats.ts`](lib/constants/siteStats.ts) both moved from `18` → `19`, propagating to the homepage, `/about`, and `/api-docs` automatically. `FREIGHTUTILS_STRATEGY.md` mcp.directory bullet updated to "live v2.1.1 / 19 tools". Historical narrative references to "11" and "18" in changelog entries, the cascade audit doc, and STATE.md PR-#45 record are intentionally LEFT — they correctly describe past or external state. CLAUDE.md "Distribution cascade" section + STATE.md distribution-surfaces table Glama row amended to encode the verified cascade behaviour: **Glama auto-publishes its build artefact on successful scrape-derived build — the "Build succeeded" maintainer email is informational, no "Create a release" click required**. Maintenance grade C → B confirmed post-publish. Added a one-liner mental-model rule: "verify Glama behaviour, do not predict it — Chrome it, do not infer it from registry semantics" (third drift in three days). Audit: [docs/audit/tool-count-standardise-2026-05-17.md](docs/audit/tool-count-standardise-2026-05-17.md). Cascade audit appended: [docs/audit/distribution-cascade-2026-05-16.md](docs/audit/distribution-cascade-2026-05-16.md) (2026-05-17 afternoon amendment). External follow-up — Soap-manual edits to `SoapyRED/freightutils-mcp` README + `server.json` (≤100 char) + npm `package.json` descriptions, and the Glama dashboard description field — listed with exact target text in the PR body.
 - **Distribution**: Glama Release-tracking gap closed. Chrome verification revealed Glama still on "Latest release: v1.0.0" + Tool Definition cache at 11 tools + Maintenance grade C — despite Sprint 1's MCP Registry cascade fix. Diagnosed that Glama scrapes the GitHub Releases API specifically; the publish workflow on [`SoapyRED/freightutils-mcp`](https://github.com/SoapyRED/freightutils-mcp) didn't cut Releases. Extended `publish-registry.yml` with a final step that creates a GitHub Release on every tag push (notes from `CHANGELOG.md`, `--latest=true` when the version matches MCP Registry isLatest); plus an idempotent guard on the publish step so re-runs on already-published versions are no-ops. Backfilled four historical Releases (v2.1.1 Latest, v2.1.0, v2.0.0, v1.1.0) via `workflow_dispatch`. Glama re-scrape lag is days, not hours — Chrome re-verify in 3–5 days. Companion PR: [SoapyRED/freightutils-mcp#3](https://github.com/SoapyRED/freightutils-mcp/pull/3) MERGED. Audit amendment: [docs/audit/distribution-cascade-2026-05-16.md](docs/audit/distribution-cascade-2026-05-16.md) (2026-05-17 section). CLAUDE.md "Distribution cascade" section amended.
 
 ## 2026-05-16

CLAUDE.md

@@ -59,7 +59,8 @@ Canonical version source for the MCP surface is **`SoapyRED/freightutils-mcp/pac
 **Days-of-lag scrape (no event link; do not block sprint on movement):**
 
 - **Smithery** — re-fetches npm + HTTP MCP endpoint on its own cadence. Score won't move same-day a release publishes. If past 7 days and still stale, request manual rescan via dashboard.
-- **Glama** — scrapes the **GitHub Releases API specifically** (not git tags, not the MCP Registry). The "Latest release" cache, Tool Definition cache, and Maintenance grade all refresh on the next Glama scrape AFTER a Release is published. Description field is dashboard-frozen (not in `glama.json`, which only accepts `maintainers`). To refresh the description, Soap edits via dashboard. The 2026-05-17 amendment to the audit doc covers the Release-API specifics — diagnosed because Sprint 1's tag-push cascade made Registry catch up but Glama remained on its first-ever scrape value.
+- **Glama** — scrapes the **GitHub Releases API specifically** (not git tags, not the MCP Registry). The "Latest release" cache, Tool Definition cache, and Maintenance grade refresh on the next Glama scrape after a Release is published. **Glama auto-publishes its build artefact when the scrape-derived build succeeds — the "Build succeeded for FreightUtils MCP Server" email is informational, no maintainer "Create a release" click required.** Maintenance grade upgrades (e.g. C → B) occur on the next scrape after the Release is cut and the auto-publish runs. Description field is dashboard-frozen (not in `glama.json`, which only accepts `maintainers`); to refresh the description Soap edits via dashboard. The 2026-05-17 amendment(s) to the audit doc cover the Release-API specifics — diagnosed because Sprint 1's tag-push cascade made Registry catch up but Glama remained on its first-ever scrape value.
+- **Glama mental-model history:** cascade was repeatedly mis-described in our docs/memory (three drifts since 2026-05-15). **Verify Glama behaviour, do not predict it** — Chrome it, do not infer it from registry semantics.
 
 **Manual UI work every release (no automation):**
 

FREIGHTUTILS_STRATEGY.md

@@ -51,7 +51,7 @@ Honest measured signals only:
 - GitHub `freightutils-mcp` — 2 stars
 - PulseMCP — listed at #3,270 global, "5.3k visitors all-time, 1.1k this week" = algorithmic estimate, not measured
 - Official MCP Registry — listed, manifest-only, no rank/install/popularity exposed
-- mcp.directory — listed but stale (lists v1.0.0 / 11 tools vs live v2.1.0 / 18 tools); the "22 reviews / 4.4 rating" was SEO JSON-LD metadata, not real user reviews
+- mcp.directory — listed but stale (lists v1.0.0 / 11 tools vs live v2.1.1 / 19 tools); the "22 reviews / 4.4 rating" was SEO JSON-LD metadata, not real user reviews
 - Glama — listed, 1 favorite (likely test/Soap)
 - Zero independent editorial coverage on Reddit, HN, dev.to, Medium, LinkedIn (beyond Marius's own posts)
 

STATE.md

@@ -141,7 +141,7 @@ Scrape-path scope (single regex): `^/(hs/(code|heading|chapter|section)|adr|airl
 | MCP Registry | **v2.1.1 (isLatest)** | published 2026-05-16T10:12Z | Workflow fires on tag push (`v*`), Release published, or `workflow_dispatch`. Publish step is now idempotent (skip if version already in Registry history). Catch-up: `gh workflow run "Publish to MCP Registry" --repo SoapyRED/freightutils-mcp --ref main`. Reference: `docs/audit/distribution-cascade-2026-05-16.md`. |
 | GitHub Releases (freightutils-mcp) | **v2.1.1 (Latest)** + v2.1.0, v2.0.0, v1.1.0 | backfilled 2026-05-17 | Cut automatically by the publish workflow's final step on every tag push (notes from `CHANGELOG.md`). v1.0.5 / v1.0.4 historical tags deferred — no CHANGELOG entries; backfill is Soap-manual if ever needed. |
 | Smithery | listed | **88/100 · 28 uses · 13 daily sessions** (pre-rescrape) | Scrape-based, days-of-lag after `freightutils-mcp` publish. No event link from MCP Registry. Rescrape on v2.1.1 pending. |
-| Glama | listed | License A · Quality A · **Maintenance C** | **Tracks the GitHub Releases API specifically.** Stale "Latest release: v1.0.0" + Tool Definition cache at 11 tools + Maintenance grade C were all downstream of the missing Releases — addressed by the 2026-05-17 release-gap sprint. Glama re-scrapes on its own cadence (days, not hours); Chrome re-verify in 3–5 days. Description still dashboard-frozen; `glama.json` schema only accepts `maintainers`. |
+| Glama | listed | License A · Quality A · **Maintenance B (post-publish)** | **Tracks the GitHub Releases API specifically.** Stale "Latest release: v1.0.0" + Tool Definition cache at 11 tools + Maintenance grade C were all downstream of the missing Releases — addressed by the 2026-05-17 release-gap sprint. **Glama auto-publishes its build artefact on successful scrape-derived build; the "Build succeeded" maintainer email is informational, no manual "Create a release" click required.** Maintenance C → B confirmed post-publish (2026-05-17). Glama re-scrapes on its own cadence (days, not hours); Chrome re-verify lag. Description still dashboard-frozen (`glama.json` schema only accepts `maintainers`) — Soap-manual dashboard edit pending. |
 | awesome-mcp-servers | listed | PR #5358 OPEN (17→19, v2.1.1) | Punkpeye-merge pending. PR description carries Glama badge + grades per April 2026 policy. |
 | Postman | live | 0 views / 0 forks | — |
 | RapidAPI | Pro live | — | — |

docs/audit/distribution-cascade-2026-05-16.md

@@ -428,3 +428,39 @@ The Sprint 1 docs covered "Smithery + Glama scrape independently" correctly. Thi
 - CLAUDE.md "Distribution cascade" section: amend the Glama bullet to "re-scrapes GitHub Releases on its own cadence."
 - `STATE.md` distribution surfaces table — Glama row: add "tracks GitHub Releases" detail.
 - `freightutils-distribution` skill (external `Skills-Claude/freightutils-distribution.md`): same amendment.
+
+---
+
+## Amendment 2026-05-17 (afternoon): Glama auto-publishes on successful build
+
+**Sprint:** `distribution-closure-tool-count-and-cascade-docs`
+**Branch:** `chore/tool-count-standardize-19-and-cascade-docs`
+
+Chrome verification ~hours after the 2026-05-17 release-gap fix landed surfaced a third correction to the cascade mental model:
+
+- **Glama auto-publishes its build artefact when the scrape-derived build succeeds.** The "Build succeeded for FreightUtils MCP Server" email Soap received is informational — no maintainer "Create a release" click required, no dashboard action needed for the build to go live. The earlier framing in CLAUDE.md / STATE.md / this audit doc implied a manual maintainer publish step gating each new build; that's wrong.
+- **Maintenance grade C → B is confirmed post-publish.** Driven by the successful Release cut + auto-publish run. The C grade was load-bearing on the missing-Releases signal alone, exactly as the 2026-05-17 (morning) amendment predicted.
+
+### Recurring lesson: verify Glama behaviour, do not predict it
+
+This is the **third time** in three days that the cascade has been mis-described in our docs/memory:
+
+| Date | Wrong claim | Verified reality |
+|---|---|---|
+| 2026-05-15 (Sprint 1) | "MCP Registry GH Action cascades to Smithery + Glama auto-sync" | Smithery + Glama are scrape-based with days-of-lag. No cascade. |
+| 2026-05-17 (morning, release-gap sprint) | "Smithery + Glama scrape independently; tag push enough" | Glama specifically scrapes the **GitHub Releases API** — tag push without Release publish is invisible to Glama. |
+| 2026-05-17 (afternoon, this amendment) | "After Release publish, maintainer cuts the Glama build via dashboard" | Glama **auto-publishes** the build artefact when its scrape-derived build succeeds. Email is informational. |
+
+CLAUDE.md "Distribution cascade" section now carries a "Glama mental-model history" one-liner: **verify Glama behaviour, do not predict it — Chrome it, do not infer it from registry semantics.**
+
+### Phase 2 of this amendment
+
+No code/workflow change required for this realisation — Glama's auto-publish is upstream of anything in our repos. The fix is documentation-only:
+
+- CLAUDE.md "Distribution cascade" section: Glama bullet updated to note the auto-publish behaviour + "email is informational" + new "verify, do not predict" mental-model line.
+- STATE.md distribution surfaces table Glama row: Maintenance grade updated `C` → `B (post-publish)`, notes amended to drop any wording implying a manual maintainer publish step, mention the auto-publish explicitly.
+
+### What's still pending Soap-manual
+
+- **Glama description dashboard edit** — description field is dashboard-frozen ("11 freight calculation and reference tools …"); `glama.json` only accepts `maintainers`. Independent of build cascade. Soap-manual when convenient.
+- **freightutils-mcp README + server.json + package.json descriptions** — out of this sandbox's repo access. External follow-up per the companion audit `docs/audit/tool-count-standardise-2026-05-17.md`.