Merge pull request #43 from Curabis/setup/bcquality-knowledge-sync

[Setup] Deploy BCQuality knowledge sync mechanism to every CURABIS project

custom/setup/curabis-standard.agent.md

@@ -1,15 +1,17 @@
 ---
 kind: action-skill
 id: curabis-standard-setup
-version: 5
+version: 6
 title: CURABIS Standard — Project Setup
 description: >
   Configures a new or existing repository to the CURABIS Standard development
   environment. Writes CLAUDE.md, BCQuality agents, .mcp.json and cspell.json
   from authoritative templates in BCQuality. Deploys bc-mcp-bridge.js to the
-  developer's machine. Also handles updates to an already-configured project.
+  developer's machine. Syncs a local three-layer BCQuality knowledge mirror
+  (custom/community/microsoft + INDEX.md) into the repo. Also handles updates
+  to an already-configured project.
 inputs: [repo-root]
-outputs: [CLAUDE.md, .mcp.json, .github/.agents/*, cspell.json, projectmemory/, docs/]
+outputs: [CLAUDE.md, .mcp.json, .github/.agents/*, .github/.agents/bcquality-knowledge/, cspell.json, projectmemory/, docs/]
 domain: setup
 keywords: [setup, bootstrap, update, mcp, bcquality, standard, new-project]
 ---
@@ -60,6 +62,7 @@ AGENTS_BASE = https://raw.githubusercontent.com/Curabis/BCQuality/main/custom/ag
 | aurelius.agent.md | `{AGENTS_BASE}/aurelius.agent.md` |
 | munger.agent.md | `{AGENTS_BASE}/munger.agent.md` |
 | cspell.json | `{BASE}/templates/cspell.json` |
+| sync-bcquality-knowledge.ps1 | `{BASE}/sync-bcquality-knowledge.ps1` |
 
 CLAUDE.md and .mcp.json are generated dynamically — not fetched as static templates
 because they contain project-specific paths.
@@ -144,15 +147,20 @@ This file is read automatically by Claude Code at the start of every session.
 At the start of every session, before doing anything else:
 
 1. Read `.github/.agents/bcquality.agent.md`
-2. Read BCQuality knowledge files from local cache (no network — fast):
+2. Read knowledge files vendored locally under `.github/.agents/bcquality-knowledge/`:
-   ```
+   - **`custom/`** — always read in full (CURABIS-specific rules, always active)
-   C:\Users\mid\.claude\bcquality-knowledge\architecture\*.md
+   - **`community/`** and **`microsoft/`** — do NOT read in full (~200 files is too
-   C:\Users\mid\.claude\bcquality-knowledge\testing\*.md
+     much to preload every session). Instead read `INDEX.md` first and open only
-   C:\Users\mid\.claude\bcquality-knowledge\mcp\*.md
+     the files whose domain/keywords match the task at hand.
-   ```
+
-   The cache is populated automatically when BCQuality updates (via global CLAUDE.md
+These files are a local copy of the Curabis BCQuality knowledge base, spanning
-   auto-update). If the cache is missing or empty on first run, the auto-update will
+all three layers (custom/community/microsoft). They are read from disk each
-   populate it. Do not fetch URLs manually unless explicitly asked.
+session — no network fetch required. Refresh them from upstream periodically
+(e.g. monthly, or whenever a session flags the mirror as stale) by running:
+
+    powershell -File .github/.agents/sync-bcquality-knowledge.ps1
+
+Then review the diff and commit any changes.
 
 These rules are always active.
 
@@ -352,6 +360,21 @@ Fetch and write verbatim:
 
 Create `.github/.agents/` if it does not exist.
 
+#### 4c-2. bcquality-knowledge/ (local mirror of all three BCQuality layers)
+
+1. Fetch `{BASE}/sync-bcquality-knowledge.ps1` → write to
+   `.github/.agents/sync-bcquality-knowledge.ps1`
+2. Run it once: `powershell -ExecutionPolicy Bypass -File .github/.agents/sync-bcquality-knowledge.ps1`
+   This populates `.github/.agents/bcquality-knowledge/{custom,community,microsoft}/`
+   plus an `INDEX.md` (domain + keywords per file, for relevance-based lookup —
+   `custom/` is always read in full, `community/` and `microsoft/` are scanned via
+   the index rather than preloaded, since together they run into the hundreds of files).
+3. Confirm: "bcquality-knowledge/ synkroniseret — [antal] filer på tværs af 3 lag."
+
+This mirror is what Step 4a's generated CLAUDE.md instructs Claude to read at
+session start. Without this step, the CLAUDE.md reference in 4a points at a
+folder that doesn't exist yet.
+
 #### 4d. cspell.json
 
 Fetch `{BASE}/templates/cspell.json` and write to repo root.
@@ -447,6 +470,8 @@ Never touches `CLAUDE.md`, `projectmemory/`, `docs/`, or `~/.bc-mcp.config.json`
 | `.github/.agents/weber.agent.md` | Fetch fresh from BCQuality, overwrite (add if missing) |
 | `.github/.agents/algo-settings.agent.md` | Fetch fresh from BCQuality, overwrite (add if missing) |
 | `.github/.agents/smiley.agent.md` | Fetch fresh from BCQuality, overwrite (add if missing) |
+| `.github/.agents/sync-bcquality-knowledge.ps1` | Fetch fresh from BCQuality, overwrite (add if missing) |
+| `.github/.agents/bcquality-knowledge/` | Re-run the sync script (see below), commit the diff |
 | `cspell.json` — words from template | Merge new words, keep project words |
 | `.mcp.json` — `al` entry | Add if `find-altool.ps1` now exists and entry is missing |
 | `.mcp.json` — `businesscentral` path | Validate and correct if wrong (see below) |
@@ -454,6 +479,27 @@ Never touches `CLAUDE.md`, `projectmemory/`, `docs/`, or `~/.bc-mcp.config.json`
 | `HEARTBEAT.md` | Create from template if missing (substitute tokens), never overwrite |
 | `docs/specs/`, `docs/decisions/`, `docs/cleanup/` | Create if missing, never overwrite content |
 
+### bcquality-knowledge/ re-sync (Mode B)
+
+After overwriting `.github/.agents/sync-bcquality-knowledge.ps1`, always re-run it:
+
+```
+powershell -ExecutionPolicy Bypass -File .github/.agents/sync-bcquality-knowledge.ps1
+```
+
+This refreshes `.github/.agents/bcquality-knowledge/{custom,community,microsoft}/`
+and regenerates `INDEX.md` from the live BCQuality tree. Run this every time Mode B
+runs, not just when the script itself changed — the mirror goes stale independently
+of the script (new upstream knowledge files land on their own schedule). Stage the
+resulting diff for the update commit in "After update — report and commit" below.
+
+If `.github/.agents/sync-bcquality-knowledge.ps1` does not exist yet (project set up
+before this mechanism existed): fetch it, run it, and also check whether the
+project's CLAUDE.md still has the old flat `architecture/*.md, testing/*.md, mcp/*.md`
+BCQuality-reading instruction from before the three-layer mirror existed. If so, propose
+replacing it with the current template from Step 4a and ask for confirmation before
+editing CLAUDE.md (same confirmation gate as the agent-synligheds-check below).
+
 ### .mcp.json — hardcoded developer-path validation (Mode B)
 
 `.mcp.json` is git-committed and shared — it must not contain a path baked in for

custom/setup/sync-bcquality-knowledge.ps1

@@ -0,0 +1,125 @@
+# Refresh the local mirror of the Curabis BCQuality knowledge base.
+#
+# Mirrors three layers from https://github.com/Curabis/BCQuality:
+#   custom/    - Curabis org-specific rules   (ALWAYS read in full each session)
+#   community/ - BC community patterns         (loaded on relevance via INDEX.md)
+#   microsoft/ - platform guardrails           (loaded on relevance via INDEX.md)
+#
+# The upstream file list is discovered dynamically from the GitHub tree API, so
+# new/removed upstream files propagate automatically - nothing is hardcoded.
+# After downloading, an INDEX.md is generated (one line per file, with domain +
+# keywords from each file's frontmatter) so an agent can scan and pull only the
+# files relevant to a task instead of loading all ~100 every session.
+#
+# Run periodically, then review the diff and commit:
+#   pwsh .github/.agents/sync-bcquality-knowledge.ps1
+
+$ErrorActionPreference = 'Stop'
+
+$repo    = 'Curabis/BCQuality'
+$branch  = 'main'
+$dest    = Join-Path $PSScriptRoot 'bcquality-knowledge'
+$staging = "$dest.tmp"
+$rawBase = "https://raw.githubusercontent.com/$repo/$branch"
+$treeUrl = "https://api.github.com/repos/$repo/git/trees/$branch" + '?recursive=1'
+
+# Upstream path prefix -> local layer folder
+$layerMap = [ordered]@{
+    'custom/knowledge/'    = 'custom'
+    'community/knowledge/' = 'community'
+    'microsoft/knowledge/' = 'microsoft'
+}
+
+function Get-Frontmatter {
+    param([string]$Path)
+    $lines = Get-Content -Path $Path
+    $fm = @{}
+    if ($lines.Count -eq 0 -or $lines[0].Trim() -ne '---') { return $fm }
+    for ($i = 1; $i -lt $lines.Count; $i++) {
+        if ($lines[$i].Trim() -eq '---') { break }
+        if ($lines[$i] -match '^\s*([\w-]+):\s*(.*)$') {
+            $fm[$matches[1]] = $matches[2].Trim()
+        }
+    }
+    return $fm
+}
+
+Write-Host "Fetching file tree from $repo@$branch ..."
+$headers = @{ 'User-Agent' = 'wareco-bcquality-sync'; 'Accept' = 'application/vnd.github+json' }
+$tree = (Invoke-RestMethod -Uri $treeUrl -Headers $headers).tree
+
+# Build the download worklist from the tree
+$files = @()
+foreach ($node in $tree) {
+    if ($node.type -ne 'blob') { continue }
+    if ($node.path -notlike '*.md') { continue }
+    foreach ($prefix in $layerMap.Keys) {
+        if ($node.path.StartsWith($prefix)) {
+            $relative = $node.path.Substring($prefix.Length)   # e.g. performance/avoid-commit-inside-loops.md
+            $files += [pscustomobject]@{
+                Url       = "$rawBase/$($node.path)"
+                Layer     = $layerMap[$prefix]
+                Relative  = $relative
+                LocalPath = Join-Path $staging (Join-Path $layerMap[$prefix] $relative)
+            }
+            break
+        }
+    }
+}
+
+if ($files.Count -eq 0) { throw 'No knowledge files found in upstream tree - aborting.' }
+
+# Download into a staging folder so a mid-run failure never wipes the live copy
+if (Test-Path $staging) { Remove-Item -Recurse -Force $staging }
+New-Item -ItemType Directory -Force $staging | Out-Null
+
+Write-Host "Downloading $($files.Count) knowledge files ..."
+$rc = 0
+foreach ($f in $files) {
+    New-Item -ItemType Directory -Force (Split-Path $f.LocalPath) | Out-Null
+    try {
+        Invoke-WebRequest -Uri $f.Url -OutFile $f.LocalPath -UseBasicParsing -ErrorAction Stop
+        Write-Host "OK   $($f.Layer)/$($f.Relative)"
+    } catch {
+        Write-Error "FAIL $($f.Layer)/$($f.Relative)"
+        $rc = 1
+    }
+}
+
+# Generate INDEX.md (relevance index for all layers)
+Write-Host "Generating INDEX.md ..."
+$idx = [System.Collections.Generic.List[string]]::new()
+$idx.Add('# BCQuality Knowledge Index')
+$idx.Add('')
+$idx.Add('<!-- Generated by sync-bcquality-knowledge.ps1 - do not edit by hand. -->')
+$idx.Add('')
+$idx.Add('Layers: `custom` is ALWAYS read in full each session. For `community` and')
+$idx.Add('`microsoft`, scan this index and read only the files whose domain/keywords')
+$idx.Add('match the task at hand.')
+$idx.Add('')
+
+foreach ($layer in @('custom', 'community', 'microsoft')) {
+    $layerFiles = $files | Where-Object { $_.Layer -eq $layer } | Sort-Object Relative
+    if (-not $layerFiles) { continue }
+    $note = if ($layer -eq 'custom') { ' (always-on)' } else { ' (load on relevance)' }
+    $idx.Add("## $layer$note")
+    $idx.Add('')
+    foreach ($f in $layerFiles) {
+        $fm = Get-Frontmatter $f.LocalPath
+        $domain = if ($fm.ContainsKey('domain')) { $fm['domain'] } else { '' }
+        $keywords = if ($fm.ContainsKey('keywords')) { $fm['keywords'].Trim('[', ']') } else { '' }
+        $rel = $f.Relative -replace '\.md$', ''
+        $idx.Add("- ``$layer/$rel`` - domain: $domain; keywords: $keywords")
+    }
+    $idx.Add('')
+}
+Set-Content -Path (Join-Path $staging 'INDEX.md') -Value $idx -Encoding utf8
+
+# Swap staging into place
+if (Test-Path $dest) { Remove-Item -Recurse -Force $dest }
+Rename-Item -Path $staging -NewName (Split-Path $dest -Leaf)
+
+Write-Host ''
+Write-Host "Done. $($files.Count) files across $($layerMap.Count) layers."
+Write-Host "Review changes with: git diff $dest"
+exit $rc