Reconcile v0.27.2 release state (#140)

* Preserve packaging migration precedence

Capture the requested packaging migration seed so the next packaging milestone surfaces the .packaging reconciliation work with the feat/package-management-pipeline branch precedence rules intact.

Constraint: .planning is ignored in this checkout, so the seed file is force-added explicitly

Constraint: Worktree already contains an unrelated unmerged .planning/phases/24-scoop-bucket-pipeline/24-01-SUMMARY.md state

Rejected: Asking for trigger and scope fields | user expects gsd-capture seed invocations to infer those fields from the command text

Confidence: high

Scope-risk: narrow

Directive: Preserve the collision rule: prefer the more complete package-manager implementation, and if both are complete prefer feat/package-management-pipeline

Tested: Read seed file; git diff --check on seed file

Not-tested: No runtime package-manager validation; this is a seed capture only

* Restore Scoop phase summary for packaging resume

The packaging milestone needs the Phase 24 execution summary to resume cleanly. The unresolved index state was deleted-by-us/updated-by-them, and the updated side preserves the completed Scoop evidence plus self-check results.

Constraint: .planning is ignored, so the tracked planning file required git add -f

Rejected: Leave the summary deleted | would remove Phase 24 completion evidence needed for packaging milestone recovery

Confidence: high

Scope-risk: narrow

Directive: Keep this summary with the packaging milestone artifacts unless Phase 24 is intentionally archived

Tested: git ls-files -u showed only this unmerged path; git diff --check on staged summary; read summary content before commit

Not-tested: No package manager runtime checks; this resolves planning state only

* Resume packaging milestone from the complete pipeline state

The package-management branch had the working release pipeline while main still carried the older .packaging tree. This reconciles the two by making packages/ the source of truth, preserving extra package managers, and completing the Nix flake publish job so the milestone resumes from Phase 26 instead of backtracking.

Constraint: Existing .packaging managers without branch equivalents had to be preserved

Constraint: GitHub release-created workflows require RELEASE_TOKEN, not GITHUB_TOKEN

Rejected: Leave .packaging as a duplicate source of truth | it would preserve stale hardcoded templates beside the pipeline templates

Confidence: medium

Scope-risk: moderate

Directive: Do not remove Chocolatey, Debian, Flatpak, or Snap templates unless a future plan explicitly drops those managers

Tested: Static workflow/template validation; deno fmt --check targeted package/docs files; git diff --check; deno task test:fast; deno task check-doc-drift; deno task test

Not-tested: deno lint remains blocked by existing repo-wide lint backlog; local nix flake check because nix is not installed; live package-manager publication and secret-backed CI jobs

* Restore the packaging milestone control surface

The implementation work was present, but the active GSD files that autonomous mode reads were missing from this checkout. This restores the v1.6 roadmap, requirements, state, and audit artifacts from the package-management worktree, then updates them to reflect the reconciled packages/ source of truth and the remaining human-runtime UAT boundary.

Constraint: .planning is ignored in this repo, so these control files were force-added intentionally

Constraint: Package-manager publication cannot be proven locally without GitHub secrets, external repos, package-manager runtimes, and a published release tag

Rejected: Mark v1.6 ready to complete | Phase verifications still contain human_needed runtime boundaries

Confidence: medium

Scope-risk: narrow

Directive: Do not run milestone completion until the v1.6 audit gaps are passed or explicitly accepted as deferred

Tested: gsd-sdk query init.milestone-op; gsd-sdk query roadmap.analyze; gsd-sdk query validate.consistency; gsd-sdk query state.validate; deno task check-doc-drift; deno task test:fast

Not-tested: deno fmt --check on GSD control markdown because Deno reflows roadmap/state fields; live AUR/Homebrew/Scoop/Winget/Nix runtime UAT

* Record implemented status for packaging reconciliation seed

The v1.6 package reconciliation already consumed this seed by migrating package-manager sources into packages/, applying the pipeline branch placeholder/checksum patterns where appropriate, and preserving auxiliary manager templates without adding live publication jobs for them. Keeping the seed dormant caused the milestone close audit to report already-consumed planning context as open work.

Constraint: .planning is ignored, so the GSD commit helper could not add the tracked seed path directly

Rejected: Leave SEED-001 dormant | it would keep audit-open blocked despite Phase 20 evidence

Confidence: high

Scope-risk: narrow

Directive: Do not reopen this seed unless future packaging work changes the migration precedence rule

Tested: gsd-sdk query audit-open --json shows seeds count 0

Not-tested: Live package-manager publication UAT

* Archive the package-manager milestone with external UAT deferred

v1.6 delivered the package-manager pipeline implementation and documentation, while the remaining package-registry proof depends on external repos, secrets, operating-system runtimes, and live release workflows. The closeout archives the milestone roadmap, requirements, audit, and phase history, collapses the active roadmap, removes the live REQUIREMENTS.md for the next cycle, and records the accepted external UAT deferrals in STATE.md.

Constraint: Homebrew and Scoop target repos returned 404, package-manager secrets were absent from gh secret list, and local nix/brew/winget runtimes are unavailable here

Rejected: Mark runtime UAT as passed | live distribution proof has not run

Rejected: Leave v1.6 open | user explicitly accepted external UAT as deferred

Confidence: high

Scope-risk: moderate

Directive: Before announcing package-manager availability, complete the deferred external UAT in STATE.md and the archived phase HUMAN-UAT files

Tested: gsd-sdk query audit-open --json

Tested: gsd-sdk query init.milestone-op

Tested: deno task check-doc-drift

Tested: git diff --cached --check

Not-tested: Live AUR/Homebrew/Scoop/Winget/Nix publication UAT

* Preserve release config while rebasing package docs

The local release-procedure commit landed on top of the merged package-manager PR with one docs conflict. The resolution keeps main's public package-manager documentation, reapplies the VitePress-safe GitHub Actions expression escaping, and keeps the .nsite release configuration changes from the original commit.

Constraint: Resolve the user's in-progress rebase onto origin/main without touching the unrelated .nsite/config.json.backup file
Rejected: Abort and restart the rebase | the remaining conflict was isolated to docs/RELEASING.md
Confidence: high
Scope-risk: narrow
Tested: deno fmt --check docs/RELEASING.md .nsite/config.json .nsite/named.json
Tested: deno task docs:build
Not-tested: full deno test suite, because this resolved docs/config rebase fallout only

* Prepare patch release above existing remote tags

The remote tag set already contains v0.27.1 while main still carried deno.json at 0.27.0, so this release advances the source-of-truth version to the next patch version before tagging.

Constraint: RELEASE.md defines deno.json as the version source of truth

Constraint: User requested the next semver from current changes with no major bump

Rejected: Minor bump to 0.28.0 | current release-candidate changes are docs/site release configuration and planning artifacts, not CLI runtime behavior

Confidence: high

Scope-risk: narrow

Directive: Do not tag a version lower than the highest existing remote semver tag

Tested: deno fmt --check deno.json

Tested: deno check src/cli.ts

Tested: deno task check-doc-drift

Tested: deno task test

Tested: deno task compile

Not-tested: Repo-wide deno fmt --check and deno lint are still failing on pre-existing broad formatting/lint debt outside this version bump

* Let package publishing find release assets without checkout

The setup job does not check out the repository, so gh release view cannot infer owner/repo. Passing github.repository explicitly keeps the asset poll working for release and manual dispatch events.

Constraint: v0.27.2 package publishing timed out despite all release assets being uploaded

Rejected: Add actions/checkout to setup | the job only needs release metadata, not the repository tree

Confidence: high

Scope-risk: narrow

Directive: Keep gh release commands repo-explicit in jobs without checkout

Tested: deno fmt --check .github/workflows/publish-packages.yml

Tested: Local asset poll against v0.27.2 found all expected standard assets

* Carry the v0.27.2 Nix flake update through PR flow

The package workflow generated the correct flake.nix changes for v0.27.2 but could not push directly to main because repository rules require pull requests. This preserves the same version and SRI hashes on the reconciliation branch.

Constraint: main rejects direct pushes under GH013 repository rules

Rejected: Rerun the same Nix job unchanged | it will regenerate the same commit and hit the same protected-branch rejection

Confidence: high

Scope-risk: narrow

Directive: Nix release updates need a PR path while main is protected

Tested: SHA256-to-SRI checks for Linux, macOS x64, and macOS arm64 release assets matched flake.nix

Not-tested: nix eval; nix is not installed in this environment

---------

Co-authored-by: sandwich <dskvr@users.noreply.github.com>

Author: dskvr

.github/workflows/publish-packages.yml

@@ -64,17 +64,19 @@ jobs:
           GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           TAG: ${{ steps.extract.outputs.tag }}
           VERSION: ${{ steps.extract.outputs.version }}
+          REPO: ${{ github.repository }}
         run: |
           TAG="${TAG}"
           VERSION="${VERSION}"
+          REPO="${REPO}"
           EXPECTED="nsyte-linux-${VERSION} nsyte-macos-arm64-${VERSION} nsyte-macos-x64-${VERSION} nsyte-windows-${VERSION}.exe"
           TIMEOUT=300
           INTERVAL=15
           ELAPSED=0
           ALL_PRESENT=false
 
           while [ "$ELAPSED" -lt "$TIMEOUT" ]; do
-            NAMES=$(gh release view "$TAG" --json assets --jq '.assets[].name' 2>/dev/null || echo "")
+            NAMES=$(gh release view "$TAG" -R "$REPO" --json assets --jq '.assets[].name' 2>/dev/null || echo "")
             ALL_PRESENT=true
             for expected in $EXPECTED; do
               if ! echo "$NAMES" | grep -qF "$expected"; then

.nsite/config.json

@@ -1,5 +1,5 @@
 {
-  "bunkerPubkey": "9c08bc9679ca77197f3e6a28876b77c5cc4b7a9dd268d9a3cffd7f1eab667cfb",
+  "bunkerPubkey": "e771af0b05c8e95fcdf6feb3500544d2fb1ccd384788e9f490bb3ee28e8ed66f",
   "fallback": "/index.html",
   "relays": [
     "wss://relay.nsite.lol",

.nsite/named.json

@@ -1,15 +1,18 @@
 {
-  "bunkerPubkey": "9c08bc9679ca77197f3e6a28876b77c5cc4b7a9dd268d9a3cffd7f1eab667cfb",
+  "bunkerPubkey": "e771af0b05c8e95fcdf6feb3500544d2fb1ccd384788e9f490bb3ee28e8ed66f",
   "relays": [
     "wss://nsite.run",
     "wss://nos.lol",
-    "wss://relay.damus.io"
+    "wss://relay.damus.io",
+    "wss://relay.nosto.re"
   ],
   "servers": [
     "https://blssm.us",
-    "https://cdn.hzrd149.com"
+    "https://cdn.hzrd149.com",
+    "https://nsite.run",
+    "https://nosto.re"
   ],
-  "id": "namednsyte",
-  "title": "nsyte (testing)",
+  "id": "nsyte",
+  "title": "nsyte.run",
   "description": "nsyte, but testing nsyte deploy to named site"
 }

.planning/MILESTONES.md

@@ -0,0 +1,69 @@
+# Milestones
+
+## v1.6 OS Package Manager/Distributors (Shipped: 2026-05-07)
+
+**Phases completed:** 7 phases, 12 plans, 8 tasks
+
+**Key accomplishments:**
+
+- Migrated package-manager sources into `packages/`, using placeholder-driven release asset templates for AUR, Homebrew, Scoop, Winget, and Nix while preserving Chocolatey, Debian, Flatpak, and Snap templates for future live-publication work.
+- Swapped release creation to `RELEASE_TOKEN` so GitHub `release: published` events can trigger downstream package publication workflows.
+- Shipped `publish-packages.yml` with asset readiness polling, centralized checksum computation, independent manager jobs, and manual per-manager `workflow_dispatch` reruns.
+- Implemented AUR, Homebrew, Scoop, Winget, and Nix publication/update jobs with fail-fast placeholder checks, idempotency guards, and runbook-backed verification.
+- Documented setup and troubleshooting for release tokens, AUR SSH, Homebrew tap PAT, Scoop bucket PAT, Winget bootstrap/update flow, and Nix flake hash updates in `docs/RELEASING.md`.
+
+**Known deferred items at close:** 6 external runtime UAT gates accepted as deferred (see `STATE.md` Deferred Items).
+
+---
+
+## Completed
+
+### v1.4: Documentation Pass
+
+**Shipped:** 2026-05-03
+**Phases:** 12-15 (12 plans, 26 tasks)
+**Goal:** Eliminate documentation drift methodically. Re-align all docs with the actual CLI surface (commands, flags, arguments, behavior). Remove "optimistic" claims for features that no longer exist or never did. Build a drift gate so this never silently regresses.
+
+**Key accomplishments:**
+
+- Rebuilt nsyte 0.26.0, captured `--help` for all 21 top-level commands + 7 bunker subcommands (29 files), and produced AUDIT-MATRIX.md joining 21 source commands × 17 doc pages × 28 help captures with severity-ordered drift classification
+- Created 5 missing command pages (get, put, scan, snapshot, status), rewrote upload.md as alias stub for deploy, re-anchored ls.md on canonical `list`, created _global-options.md as canonical globals reference
+- Aligned 7 command pages (announce, bunker, debug, download, init, serve, validate) flag-by-flag with source `.option(...)` declarations; updated commands.md index; ZERO drift across all 21 commands
+- Reconciled 8 guide pages: replaced phantom `appHandler.enabled` with `publishAppHandler`, removed phantom env vars (`NSYTE_BUNKER_KEY`, `NSYTE_LOG_LEVEL`), fixed Linux/Windows storage-path drift, deleted 227-line commented Package Managers block, updated stale `@0.10.1` pin
+- Updated 5 cross-reference pages (docs/index, README, mkdocs.yml, docs/usage/index, docs/usage/configuration); mkdocs strict build clean
+- Shipped `scripts/check-doc-drift.ts` (3 checks: coverage, flag alignment, env-var) and `.github/workflows/doc-drift.yml` CI gate
+- `v1.4-PHANTOM-LOG.md` accumulates 34 entries documenting every removed/rewritten claim with verbatim quotes — backlog for future feature work
+
+### v1.3: Redesign Homepage
+
+**Shipped:** 2026-05-03
+**Phases:** 10-11 (6 plans, 18 tasks)
+**Goal:** Compress the marketing page into a tight, above-the-fold landing with darker chrome and the existing green accent, while keeping deeper docs reachable via links.
+
+**Key accomplishments:**
+
+- Locked near-black palette (`--cyber-bg #06070d`) and tightened density scale (15px body / 1.5 line-height / 2rem section padding / ~15% reduced heading clamps) on `website/src/index.html`
+- Removed four below-the-fold sections (features, what's-nsite, CI/CD, demo walkthrough), animated grid, animated gradient, and asciinema demo player — file shrank from 1,775 → 653 lines (-63%)
+- Built top-link nav (Docs · GitHub · Install · Releases) and rebuilt hero around wordmark + tagline + particles only
+- Compact install component with underlined tabs (script default), single canonical command per tab, copy-to-clipboard + 1.5s feedback, "More install options →" link to docs
+- Folded all Phase 10 UI-audit color findings into Phase 11 rebuild (no purple/blue gradients on UI surfaces; legacy neon green replaced; Phase 11 audit scored 24/24)
+- User-approved cycle-2 layout: content flex-centered both axes (`min-height: 100vh` restored), nav moved to last child of `.hero-content`, `history.pushState` keeps URL hash in sync on `#install` clicks
+- A11y: `:focus-visible` rings, `prefers-reduced-motion` block, full ARIA tablist semantics
+
+### v1.2: Custom Event Timestamps
+
+**Shipped:** 2026-04-14
+**Phases:** 7-9
+**Goal:** Add CLI argument to override `created_at` timestamp on nostr events, with tests across all code paths.
+
+### v1.1: Address Open Issues
+
+**Shipped:** 2026-04-04
+**Phases:** 3-6
+**Goal:** Fix manifest title/description bug, add dry-run mode, config schema hosting, and secrets scanner (#100-#103).
+
+### v1.0: Test Coverage Low-Hanging Fruit
+
+**Shipped:** 2026-03-27
+**Phases:** 1-2
+**Goal:** Increase line coverage from 35.3% by testing pure-function and utility modules.

.planning/PROJECT.md

@@ -0,0 +1,134 @@
+# nsyte
+
+## What This Is
+
+A CLI tool for publishing static sites to nostr relays and blossom servers. Users can deploy, manage, and serve websites using the nostr protocol, with support for multiple named sites, snapshots, NIP-89 app handlers, and bunker-based signing. The marketing site at `nsyte.run` introduces the project with a compact above-the-fold landing that links to the full documentation site.
+
+## Core Value
+
+Reliable site deployment to the nostr + blossom ecosystem with a simple CLI workflow.
+
+## Current State
+
+**Shipped:** v1.6 OS Package Manager/Distributors on 2026-05-07.
+
+**Current package-management state:** Package templates live under `packages/`, release-triggered package jobs are wired in `.github/workflows/publish-packages.yml`, and maintainer runbooks live in `docs/RELEASING.md`. Live package-manager publication UAT remains deferred until external repos, secrets, package-manager runtimes, and release runs are available.
+
+## Requirements
+
+### Validated
+
+<!-- Shipped and confirmed valuable. -->
+
+- Site deployment to blossom servers and nostr relays (deploy, undeploy) — v1.0
+- Named site management (sites command) — v1.0
+- File operations (put, get, delete, list, download) — v1.0
+- Manifest snapshots for versioning — v1.0
+- Configuration management with bunker-based signing — v1.0
+- NIP-89 app handler announcements — v1.0
+- CI/CD integration mode — v1.0
+- Debug and validation tooling — v1.0
+- Manifest title/description event metadata fix — v1.1
+- Dry-run mode for deploy — v1.1
+- Config schema hosting — v1.1
+- Secrets scanning pre-deploy check — v1.1
+- Custom `--created-at` global CLI option for overriding event timestamps — v1.2
+- Above-the-fold homepage layout (heading, links, compact install box) — v1.3
+- Darker dark-theme palette (`#06070d` base) preserving the green accent — v1.3
+- Tighter typography, spacing, and component density on the homepage — v1.3
+- Persistent links from homepage to the existing docs site — v1.3
+- Documentation aligned 1:1 with the actual CLI surface (commands, flags, behavior) — v1.4
+- Drift gate (`scripts/check-doc-drift.ts` + GitHub Actions workflow) catches future drift in CI — v1.4
+- VitePress docs/site migration with homepage-matched theme and local search — v1.5
+- Package source-of-truth under `packages/` with AUR, Homebrew, Scoop, Winget, and Nix templates — v1.6
+- Release-triggered package publication workflow with independent manager jobs and shared asset/checksum setup — v1.6
+- Package-manager maintainer runbooks for tokens, repo bootstrap, local verification, and troubleshooting — v1.6
+
+### Active
+
+<!-- Current scope. Building toward these. -->
+
+Next milestone not selected yet. Start with `$gsd-new-milestone`.
+
+Backlog still standing from v1.4 phantom log: env-var override support, `nsyte config show`/`set` subcommands, `validate` redundant `-f` flag fix, `deploy` duplicate `-c` short-flag fix.
+
+### Out of Scope
+
+<!-- Explicit boundaries. Includes reasoning to prevent re-adding. -->
+
+- Light theme — dark-only by design (cypherpunk aesthetic)
+- Mobile-first homepage redesign — desktop above-the-fold fit is the priority; mobile gets graceful degradation only
+- Asciinema/demo content on the homepage — demo lives in docs
+- Marketing content on homepage (features, CI/CD, "what's nsite") — migrated to docs to enforce no-scroll constraint
+- CLI source changes during the v1.4 docs pass — if a doc audit reveals a real source bug or missing feature, it's flagged for a future milestone, not fixed in this one
+- CLI source changes during v1.5 — same rule: docs migration only; phantom-log backlog stays deferred
+- Live Chocolatey, Debian, Flatpak, and Snap publication — templates are preserved under `packages/`, but v1.6 intentionally did not add live jobs for those managers
+
+## Context
+
+- **Runtime:** Deno (TypeScript)
+- **Nostr libraries:** applesauce-core, applesauce-signers, applesauce-relay
+- **CLI framework:** Cliffy (Command, Prompt, Ansi)
+- **Event signing:** ISigner interface from applesauce-signers (supports bunker, nsec)
+- **Event timestamps:** `--created-at` global option overrides timestamp on most event kinds; delete (kind 5) and blossom auth (kind 24242) events always use current time
+- **Commands:** init, deploy, list, status, snapshot, sites, browse, download, get, ci, put, run, serve, delete, undeploy, validate, debug, announce, config, bunker, scan
+- **Marketing site:** Single-page HTML at `website/src/index.html` (~653 lines after v1.3 redesign, down from ~1890). Cypherpunk theme, `--cyber-bg #06070d`, `--cyber-accent #34d399`. Single `<section class="hero">` with flex-centered content (wordmark + tagline + compact install component + top-link nav). UI-audit score 24/24 at v1.3 close.
+- **Docs site:** VitePress build under `docs/`, with palette/typography matched to the homepage. Homepage links to preserve: `/docs/installation/`, `/docs/`.
+- **Package managers:** `packages/` is the package source-of-truth. AUR, Homebrew, Scoop, Winget, and Nix have v1.6 publication/update automation; Chocolatey, Debian, Flatpak, and Snap templates are preserved without live jobs.
+
+## Constraints
+
+- **Deno:** Must compile to single binary for Linux, macOS, Windows
+- **Nostr protocol:** Events must conform to NIP specifications
+- **Blossom protocol:** Auth events (kind 24242) have expiration requirements
+- **Brand color:** Existing green accent (`--cyber-accent #34d399`) is fixed — only darks deepen
+- **Marketing site:** Single-file inline CSS+HTML+JS (`website/src/index.html`); no build step
+
+## Key Decisions
+
+| Decision | Rationale | Outcome |
+|----------|-----------|---------|
+| Use applesauce libraries | Modern nostr library ecosystem with good Deno support | -- Pending |
+| Cliffy for CLI | Mature Deno CLI framework | -- Pending |
+| `--created-at` excluded from delete & blossom auth events | Deletion timestamp must reflect actual time; blossom expiration math depends on current time | ✓ Good (v1.2) |
+| Homepage flex-centered + `min-height: 100vh` (v1.3 cycle-2) | First attempt dropped 100vh and left content top-anchored — user feedback restored centering as the more pleasant layout | ✓ Good (v1.3) |
+| Nav placed AT END of `.hero-content` (v1.3 cycle-2) | User checkpoint feedback: links sit better below the wordmark/tagline/install | ✓ Good (v1.3) |
+| Underlined text tabs over button-pill style (v1.3) | Denser, fits the "tighter" mandate; cleaner with the compact install card | ✓ Good (v1.3) |
+| Phase 10 audit findings folded into Phase 11 rebuild | The flagged surfaces (install/code/copy components) were already being rebuilt; one-pass fix preferred over separate cleanup phase | ✓ Good (v1.3) — clean 24/24 audit |
+| `history.pushState` on in-page anchor click (v1.3) | Plan must_haves required `window.location.hash === '#install'`; default `e.preventDefault()` + `scrollIntoView` doesn't update the URL | ✓ Good (v1.3) |
+
+## Evolution
+
+This document evolves at phase transitions and milestone boundaries.
+
+**After each phase transition** (via `/gsd:transition`):
+1. Requirements invalidated? -> Move to Out of Scope with reason
+2. Requirements validated? -> Move to Validated with phase reference
+3. New requirements emerged? -> Add to Active
+4. Decisions to log? -> Add to Key Decisions
+5. "What This Is" still accurate? -> Update if drifted
+
+**After each milestone** (via `/gsd:complete-milestone`):
+1. Full review of all sections
+2. Core Value check -- still the right priority?
+3. Audit Out of Scope -- reasons still valid?
+4. Update Context with current state
+
+## Key Decisions
+
+(continuing from above table)
+
+| Decision | Rationale | Outcome |
+|----------|-----------|---------|
+| Phantom log as living artifact | Records every removed doc claim with verbatim quotes — future feature backlog source | ✓ Good (v1.4) — 34 entries surface concrete source-side work |
+| `upload.md` kept as alias stub (not deleted) | UX: a user searching "upload" should find a doc that points them to deploy | ✓ Good (v1.4) |
+| `ls.md` filename retained, content re-anchored on `list` | Avoid breaking external links to `/docs/usage/commands/ls/` | ✓ Good (v1.4) |
+| `_global-options.md` canonical reference | Don't duplicate `--config`/`--created-at` across 22 pages | ✓ Good (v1.4) |
+| Drift gate: 3 checks (coverage, flag, env-var) | Maps directly to the top-3 phantom patterns from the audit | ✓ Good (v1.4) |
+| Phase 14 audit folded color findings from Phase 11 audit (UI carry-over) — n/a in v1.4 (docs-only) | — | — |
+| `packages/` as package source-of-truth | Keeps manager templates versioned with the release workflow and avoids hidden `.packaging` drift | ✓ Good (v1.6) |
+| Independent package manager jobs instead of a matrix | One package-manager failure should not cancel unrelated publication targets | ✓ Good (v1.6) |
+| External package UAT deferred at close | Live pushes require repos, secrets, package-manager runtimes, and package index workflows outside local static verification | ⚠ Revisit before public package announcement |
+
+---
+*Last updated: 2026-05-07 — after v1.6 milestone close*

.planning/RETROSPECTIVE.md

@@ -0,0 +1,57 @@
+# Project Retrospective
+
+*A living document updated after each milestone. Lessons feed forward into future planning.*
+
+## Milestone: v1.6 - OS Package Manager/Distributors
+
+**Shipped:** 2026-05-07
+**Phases:** 7 | **Plans:** 12
+
+### What Was Built
+
+- Package-manager source templates moved into `packages/` for AUR, Homebrew, Scoop, Winget, Nix, Chocolatey, Debian, Flatpak, and Snap.
+- `publish-packages.yml` wires release-triggered package publication with shared asset polling/checksum setup and independent manager jobs.
+- AUR, Homebrew, Scoop, Winget, and Nix publication/update paths now have CI implementations and `docs/RELEASING.md` runbooks.
+
+### What Worked
+
+- Preserving `.packaging` analogs before applying the pipeline branch patterns avoided losing Chocolatey, Debian, Flatpak, and Snap coverage.
+- Independent package-manager jobs matched the failure-isolation goal and kept each manager's validation readable.
+- The final open-artifact audit caught stale debug sessions and an already-consumed seed before archive.
+
+### What Was Inefficient
+
+- External package-manager UAT could not close locally because the required GitHub repos, secrets, package-manager runtimes, and live release workflows were not available.
+- Some generated milestone accomplishments needed manual cleanup before they were useful as a durable summary.
+
+### Patterns Established
+
+- `packages/` is the package source-of-truth.
+- Placeholder replacement must be named, fail-fast, and idempotent for package templates.
+- Runtime distribution proof should be split from static pipeline proof when external registries and secrets are the real gate.
+
+### Key Lessons
+
+1. Check external repo and secret existence early in package-manager milestones; missing infrastructure should be recorded as a live-UAT boundary before close.
+2. Treat dormant seeds as close blockers only if their trigger remains unimplemented; otherwise mark them implemented with phase evidence.
+3. Archive package-manager milestones with explicit external-UAT deferrals so future release work has a precise checklist rather than a vague "human needed" state.
+
+### Cost Observations
+
+- Sessions: multiple interrupted/resumed GSD passes.
+- Notable: targeted Deno gates plus `audit-open` were more useful than repo-wide lint for closeout confidence.
+
+---
+
+## Cross-Milestone Trends
+
+### Process Evolution
+
+| Milestone | Phases | Key Change |
+|-----------|--------|------------|
+| v1.6 | 7 | Package-manager work needs static implementation gates plus separate external runtime gates. |
+
+### Top Lessons
+
+1. Keep durable planning artifacts, but close or archive stale open-state records before milestone completion.
+2. Prefer repo-local validation for implementation proof and make external live-UAT prerequisites concrete.