Add Claude Code project configuration and memory

.claude/skills/cv-check/SKILL.md

@@ -0,0 +1,53 @@
+---
+name: cv-check
+description: Render the CV with rendercv and check if the output PDF fits within 1 page. If it exceeds 1 page, analyze the YAML content and make it more compact — shorten summaries, simplify highlights, remove redundant or low-value information — until it fits. Use this whenever the user asks to check, validate, or review the CV, or wants to know if it fits on one page.
+---
+
+# CV Check
+
+Render the CV and verify it fits within 1 page. If not, compact the YAML content until it does.
+
+## Step 1: Render
+
+```bash
+cd "D:/tiennm99/cv" && PYTHONUTF8=1 .venv/Scripts/rendercv render miti99.yml
+```
+
+`PYTHONUTF8=1` is required on Windows to avoid encoding errors with rendercv's rich output.
+
+## Step 2: Count pages
+
+```bash
+cd "D:/tiennm99/cv" && .venv/Scripts/python.exe -c "
+from pypdf import PdfReader
+reader = PdfReader('miti99.pdf')
+print(len(reader.pages))
+"
+```
+
+If `pypdf` is missing: `.venv/Scripts/pip install pypdf -q`
+
+## Step 3: Evaluate
+
+- **1 page** → PASS. Tell the user the CV renders cleanly to 1 page.
+- **>1 page** → FAIL. Proceed to Step 4.
+
+## Step 4: Compact the YAML (if > 1 page)
+
+The fix is always in `miti99.yml` — never manipulate the PDF directly.
+
+Priorities for compacting, in order:
+
+1. **Shorten summaries** — experience and project summaries are the biggest space consumers. Trim to 1–2 tight sentences, remove filler phrases.
+2. **Compress highlights** — merge bullets that cover similar points, cut anything that doesn't add new information beyond the summary.
+3. **Remove low-value projects** — if there are 4+ projects, consider dropping the weakest one entirely.
+4. **Collapse education highlights** — if GPA/awards take multiple lines, merge into one.
+5. **Shorten skills** — remove categories that are implied or redundant.
+
+After editing, re-run from Step 1 and repeat until the CV passes.
+
+## Guiding principles
+
+- Preserve all key technical facts, metrics (CCU, %, team size), and named technologies.
+- Prefer cutting vague or soft language ("passionate about", "various tools") before cutting substance.
+- Never fabricate or add new information.