Thanks for the careful work here, @AnandSundar — this is a clean, well-scoped metadata PR that hits all six items from #3675. I verified the substance against the repo rather than just the AppStream schema:

Verified accurate ✅

• Feature highlights — every <li> maps to a real README claim: native SmartSDR protocol v1.4.0.0, GPU spectrum + multi-panadapter + native VITA-49, Aetherial Audio Channel Strip with six NR engines, Multi-Flex + SmartLink + TCI v2.0, AetherModem and FreeDV RADE. No invented features.
• Releases — all five versions correspond to real upstream tags (v26.6.3 / v26.6.2 / v26.6.1.1 / v26.6.1 / v26.5.3), and the dates I could check against local tags (26.6.3 → 2026-06-14, 26.6.1 → 2026-06-01, 26.5.3 → 2026-05-24) match exactly.
• Screenshot — docs/assets/screenshot-current.png does exist at the pinned v26.6.3 tag, so swapping the moving refs/heads/main ref for a tagged one is both correct and resolves the moving-target concern from the issue.

Minor — non-blocking

• Missing trailing newline — the file now ends without a final newline (\ No newline at end of file in the diff). Worth adding one to keep POSIX-text linters / pre-commit hooks happy.
• <developer> divergence from the issue — #3675 suggested id="github.com/aethersdr" / name "The AetherSDR Project"; you used id="io.github.aethersdr" / "AetherSDR". Just flagging the difference — the reverse-DNS id form you chose is actually the more current AppStream convention, so no change needed unless the maintainers want the longer project name displayed.

Optional thought for later: the <release> entries have no <description>/<url>, so software centres will show empty release notes — a url pointing at each GitHub release tag would surface changelogs, but that's an enhancement, not a gap in this PR.

Nice job pinning everything to citable sources. LGTM from a correctness standpoint.

_{🤖 aethersdr-agent · cost: $3.4456 · model: claude-opus-4-8}

Human-directed independent review (Claude / @NF0T) | Request Changes

This review was conducted by Claude under direction from @NF0T, independently of the triage bot's prior pass. Every factual claim in the PR was verified against the current codebase, git history, and GitHub's release API before forming a conclusion. A maintenance sustainability concern identified by @NF0T during review has also been folded in below.

What I verified and how

PR description accuracy — one overclaim identified.

I read the diff against the pre-PR context lines to establish which elements were actually changed. The <launchable type="desktop-id">AetherSDR.desktop</launchable> element appears in the diff as unchanged context (no + or −). It was already present before this commit and is not touched by it. The PR description's claim that it "bound type=\"desktop-id\" to AetherSDR.desktop" is incorrect — this element pre-existed. The six remaining claimed changes are real and confirmed in the diff.

Release versions and dates — verified against git and GitHub release API.

The bot review noted it could check only three of the five release dates against local tags. I checked all five against GitHub's release API directly:

The v26.6.2 tag was not in the local repo after pulling upstream — I verified its existence via the GitHub refs API (object SHA 2629aa79). The one-day offsets for v26.6.2 and v26.6.1.1 are consistent with all five dates being expressed in PDT (UTC−7): a tag created at 20:37 PDT on June 1 is June 2 in UTC, which is what GitHub's API reports. All five dates appear to use the tagger's local timezone consistently — a legitimate convention.

Feature highlight accuracy — verified against README.

Every <li> in the highlights block maps to a real README claim. No invented features. The Channel Strip and six NR engines are two adjacent README bullets condensed into one — the substance is accurate.

Screenshot URL, ancillary URLs — verified.

docs/assets/screenshot-current.png exists in the working tree; tag v26.6.3 confirmed via release API. SUPPORT.md exists at repo root. README.md contains ## Contributing, which resolves to the #contributing anchor. The <launchable> binding AetherSDR.desktop is consistent with the CMakeLists.txt install rule at line 2535 — the generated desktop file is installed under that name. The reverse-DNS naming Flatpak conventionally expects is a pre-existing issue, not introduced here.

Blocking issue — missing trailing newline

The diff ends with \ No newline at end of file. The file is not a valid POSIX text file in this state. Pre-commit hooks that enforce POSIX text compliance will flag it; some XML parsers and packaging tools behave unexpectedly on files without a final newline. The bot noted this as non-blocking — I disagree: it is trivial to fix and should be fixed before a file destined for a Flathub submission lands in main.

Human-identified concern — maintenance sustainability

@NF0T raised this during review and asked that it be included in scope.

This PR hardcodes both the <releases> block and the screenshot URL to a specific point in time. As written, every future release requires a manual PR to prepend a new <release> entry, trim the list, and update the screenshot URL to the new tag — and whoever cuts the release has to remember to do it. That is process debt that will compound with every release cadence.

Notably, issue #3675 already anticipated this: "Could later be auto-generated from tags in CI." The project is on record knowing the current state is incomplete.

The right long-term fix is a workflow triggered on v* tag push that: (1) prepends the new <release version="…" date="…"/> entry and trims to the last N; (2) updates the screenshot URL to the new tag; and (3) commits the result back or opens a PR. This is standard practice for AppStream-enabled projects targeting Flathub.

This concern does not block the current PR — what is here is correct for the immediate submission goal and materially better than what came before. A follow-up issue has been filed to track the CI automation work.

Summary

Approving. Re-verified the amended branch (ad8cfad8):

• NF0T's blocking item is fixed — the file now ends with </component>\n (confirmed via od -c); valid POSIX text. XML well-formed, CI green on all six checks.
• Substance triple-verified (bot, NF0T against the release API, and me): all five <release> versions/dates real and internally consistent (PDT), every feature <li> maps to a real README claim, screenshot pinned to the published v26.6.3 artifact, <launchable> binding matches the CMakeLists install rule.
• PR-body overclaim corrected (the pre-existing <launchable> was removed from the changes list).
• Maintenance concern deferred correctly to #3718 (CI auto-update of the releases block + screenshot URL on v* tag) — non-blocking, as #3675 anticipated.

Thanks @NF0T for the thorough independent pass and @AnandSundar for the quick, well-documented turnaround. Landing since the one blocking item is resolved.

(Process note for later, per @AnandSundar: AGENTS.md's "assign yourself before PR interaction" step can't be performed by fork contributors — the assign API 403s without repo write access. Worth a one-line clarification.)