@ngocsangairvds/vsaf 4.2.0 → 4.2.2

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ngocsangairvds/vsaf",
-  "version": "4.2.0",
+  "version": "4.2.2",
   "description": "logging step",
   "main": "packages/core/dist/index.js",
   "types": "packages/core/dist/index.d.ts",

package/skills/sdlc/onboard-code/SKILL.md

@@ -62,12 +62,17 @@ Icon: `⏳` running, `✅` done, `❌` error.
 [ONBOARD-CODE] [2/6] GitNexus index... ⏳
 ```
 
-3. For EACH repo found:
-   - Run `npx gitnexus analyze {path}` (if not already indexed)
-     - ⚠️ Use `npx gitnexus` (resolves local binary) — do NOT use `npx -y gitnexus@latest` (native module crash on Node 24+)
-     - If project has NO `.git/` → add `--skip-git` flag: `npx gitnexus analyze {path} --skip-git`
+3. Run GitNexus analyze at **project root** (NOT per sub-repo):
+   ```
+   Is there .git at project root?
+   ├── YES → npx gitnexus analyze .
+   └── NO  → npx gitnexus analyze . --skip-git
+   ```
+   - ⚠️ Use `npx gitnexus` (resolves local binary) — do NOT use `npx -y gitnexus@latest` (native module crash on Node 24+)
+   - This creates a SINGLE `.gitnexus/` at project root covering ALL code (including sub-repos)
+   - For multi-repo: one unified index enables cross-service analysis naturally
    - Run GitNexus `group_sync` to index
-4. Map structure per repo: `route_map` + `tool_map` + `context` (3-5 key entry points)
+4. Map structure: `route_map` + `tool_map` + `context` (3-5 key entry points per repo)
 
 ```
 [ONBOARD-CODE] [2/6] GitNexus index... ✅
@@ -81,12 +86,12 @@ Icon: `⏳` running, `✅` done, `❌` error.
 
 **IMPORTANT:** Claude MUST review GitNexus index — DO NOT trust blindly. Verify using MCP tools:
 
-5. `list_repos` — confirm all repos have been indexed
-6. `route_map` — check that API routes have been fully mapped
-7. `context` — query 3-5 key entry points, verify results make sense:
-   - Are main entry points indexed?
+5. `list_repos` — confirm project root has been indexed (single unified index)
+6. `route_map` — check that API routes from ALL repos have been fully mapped
+7. `context` — query 3-5 key entry points per repo, verify results make sense:
+   - Are main entry points indexed for each repo?
    - Are dependencies between modules correct?
-   - If multi-repo: are cross-service calls captured?
+   - If multi-repo: are cross-service calls captured? (HTTP, gRPC, event-driven)
 8. If missing:
    - Re-run `group_sync` with different config
    - Or record note in env-setup.md: "{module/route} not indexed — needs manual review"
@@ -175,7 +180,7 @@ Print ONBOARD-CODE REPORT:
 
 - `.vsaf/docs/onboarding/env-setup.md` — project structure + tech stack + env check + Docker
 - `CONTEXT.md` — updated with technical terms
-- GitNexus index ready for every repo
+- Single GitNexus index at project root (`.gitnexus/`) covering all code
 
 ## Artifact Auto-Save
 

package/skills/sdlc/onboard-docs/SKILL.md

@@ -44,13 +44,18 @@ Icon: `⏳` running, `✅` done, `❌` error.
    # Windows: "${VENVS}\markitdown\Scripts\python.exe"
    # Fallback: ~/.local/pipx/venvs/markitdown/bin/python
    ```
-   If markitdown Python not found, skip markitdown conversion — files will be classified and routed through tier fallbacks (graphify-direct-read for Tier 1, Claude Read for Tier 2 HTML).
+   If markitdown Python not found, skip step 5 entirely and route ALL complex files through tier fallbacks:
+   - **All Tier 1 files** → add to graphify-direct-read list (graphify reads natively)
+   - **All Tier 2 HTML files** → Claude Read fallback (step 6)
+   - **All Tier 2 PPTX/EPUB/ZIP files** → skipped (no fallback available)
+   Then jump to step 6 routing.
+
 4. Classify each complex file into a tier:
    - **Tier 1 (graphify-native):** PDF, DOCX, XLSX, PNG, JPG, GIF, WEBP — graphify can read these directly if markitdown fails
    - **Tier 2 (graphify-blind):** HTML, PPTX, EPUB, ZIP — graphify cannot read these; need markitdown or Claude fallback
    - Text files (`.md`, `.txt`) are NOT complex — read directly, skip conversion
 
-5. For each complex file, attempt markitdown conversion:
+5. For each complex file, attempt markitdown conversion (skip if markitdown not found — see step 3):
    - Resolve markitdown Python (step 3 above)
    - Convert using the wrapper script (bundled in this skill directory):
      ```bash