feat: Jake's Resume template as default + /add command

.agents/skills/career-ops/SKILL.md

@@ -3,7 +3,7 @@ name: career-ops
 description: AI job search command center -- evaluate offers, generate CVs, scan portals, track applications
 arguments: mode # Claude Code specific
 user-invocable: true
-argument-hint: "[scan | deep | pdf | oferta | ofertas | apply | batch | tracker | pipeline | contacto | training | project | interview-prep | update]"
+argument-hint: "[scan | deep | pdf | oferta | ofertas | apply | batch | tracker | pipeline | contacto | training | project | interview-prep | add | update]"
 license: MIT
 ---
 
@@ -32,6 +32,7 @@ Determine the mode from `$mode`:
 | `batch` | `batch` |
 | `patterns` | `patterns` |
 | `followup` | `followup` |
+| `add` | `add` |
 
 **Auto-pipeline detection:** If `$mode` is not a known sub-command AND contains JD text (keywords: "responsibilities", "requirements", "qualifications", "about the role", "we're looking for", company name + role) or a URL to a JD, execute `auto-pipeline`.
 
@@ -63,6 +64,7 @@ Available commands:
   /career-ops batch     → Batch processing with parallel workers
   /career-ops patterns  → Analyze rejection patterns and improve targeting
   /career-ops followup  → Follow-up cadence tracker: flag overdue, generate drafts
+  /career-ops add {url|text} → Add a project, experience, or skill to your CV permanently
 
 Inbox: add URLs to data/pipeline.md → /career-ops pipeline
 Or paste a JD directly to run the full pipeline.
@@ -82,7 +84,7 @@ Applies to: `auto-pipeline`, `oferta`, `ofertas`, `pdf`, `contacto`, `apply`, `p
 ### Standalone modes (only their mode file):
 Read `modes/{mode}.md`
 
-Applies to: `tracker`, `deep`, `interview-prep`, `training`, `project`, `patterns`, `followup`
+Applies to: `tracker`, `deep`, `interview-prep`, `training`, `project`, `patterns`, `followup`, `add`
 
 ### Modes delegated to subagent:
 For `scan`, `apply` (with Playwright), and `pipeline` (3+ URLs): launch as Agent with the content of `_shared.md` + `modes/{mode}.md` injected into the subagent prompt.

AGENTS.md

@@ -210,6 +210,7 @@ Default modes are in `modes/` (English). Additional language-specific modes are
 | Batch processes offers | `batch` |
 | Asks about rejection patterns or wants to improve targeting | `patterns` |
 | Asks about follow-ups or application cadence | `followup` |
+| Wants to add a project, GitHub repo, or experience to their CV | `add` |
 
 ### CV Source of Truth
 

config/profile.example.yml

@@ -67,8 +67,10 @@ location:
   # onsite_availability: "1 week/month in any city"
 
 cv:
-  output_format: "html" # "html" (default) or "latex"
+  # CV output format. Default is "latex" (Jake's Resume template, compiled via tectonic/pdflatex).
+  # Set to "html" to use the HTML/Playwright PDF pipeline instead.
+  output_format: "latex"
   # (Optional) Canva resume design ID for visual CV generation via /career-ops pdf.
   # Find it in your Canva design URL: https://www.canva.com/design/DAxxxxxxx/...
   # The ID starts with "D" and is 11 characters long.
-  # canva_resume_design_id: "DAxxxxxxxxx"
+  # canva_resume_design_id: "DAxxxxxxxxx"

generate-latex.mjs

@@ -20,8 +20,8 @@ import { existsSync, mkdirSync } from 'fs';
 
 const REQUIRED_SECTIONS = [
   '\\\\section{Education}',
-  '\\\\section{Work Experience}',
-  '\\\\section{Personal Projects}',
+  '\\\\section{Experience}',
+  '\\\\section{Projects}',
   '\\\\section{Technical Skills}',
 ];
 
@@ -90,9 +90,9 @@ async function main() {
     if (/\\resumeProjectHeading/.test(line)) projectHeadingCount++;
   }
 
-  // Check pdfgentounicode
+  // Check pdfgentounicode (accept bare form or wrapped in \ifpdf)
   if (!content.includes('\\pdfgentounicode=1')) {
-    issues.push('Missing \\pdfgentounicode=1 (ATS compatibility)');
+    issues.push('Missing \\pdfgentounicode=1 (ATS compatibility — wrap in \\ifpdf...\\fi for tectonic/XeLaTeX compatibility)');
   }
 
   const fileInfo = await stat(absPath);
@@ -177,7 +177,7 @@ async function main() {
       ];
       // First pass
       execFileSync('pdflatex', pdflatexArgs, { cwd: texDir, stdio: 'pipe', timeout: 120_000 });
-      // Second pass (resolves referenceS))
+      // Second pass (resolves references)
       execFileSync('pdflatex', pdflatexArgs, { cwd: texDir, stdio: 'pipe', timeout: 120_000 });
     }
 

modes/_shared.md

@@ -122,11 +122,11 @@ After detecting archetype, read `modes/_profile.md` for the user's specific fram
 | WebSearch | Comp research, trends, company culture, LinkedIn contacts, fallback for JDs |
 | WebFetch | Fallback for extracting JDs from static pages |
 | Playwright | Verify offers (browser_navigate + browser_snapshot). **NEVER 2+ agents with Playwright in parallel.** |
-| Read | cv.md, _profile.md, article-digest.md, cv-template.html |
-| Write | Temporary HTML for PDF, applications.md, reports .md |
+| Read | cv.md, _profile.md, article-digest.md, cv-template.tex |
+| Write | Generated .tex CV, applications.md, reports .md |
 | Edit | Update tracker |
 | Canva MCP | Optional visual CV generation. Duplicate base design, edit text, export PDF. Requires `cv.canva_resume_design_id` in profile.yml. |
-| Bash | `node generate-pdf.mjs` |
+| Bash | `node generate-latex.mjs` (default); `node generate-pdf.mjs` if `cv.output_format: html` |
 
 ### Time-to-offer priority
 - Working demo + metrics > perfection
@@ -226,7 +226,7 @@ These rules apply to ALL generated text that ends up in candidate-facing documen
 - "demonstrated ability to" / "best practices" (name the practice)
 
 ### Unicode normalization for ATS
-`generate-pdf.mjs` automatically normalizes em-dashes, smart quotes, and zero-width characters to ASCII equivalents for maximum ATS compatibility. But avoid generating them in the first place.
+`generate-latex.mjs` compiles the `.tex` file to PDF via `pdflatex` or `tectonic` (auto-detected). For the HTML path, `generate-pdf.mjs` normalizes em-dashes, smart quotes, and zero-width characters to ASCII. Either way, avoid generating non-ASCII characters in the first place.
 
 ### Vary sentence structure
 - Don't start every bullet with the same verb

modes/add.md

@@ -0,0 +1,193 @@
+# Mode: add — Add New Content to CV
+
+Add a project, experience, publication, or skill to `cv.md` and `article-digest.md` so every future CV generation includes it.
+
+## Input Formats
+
+The user can pass:
+- **GitHub URL** — `https://github.com/user/repo`
+- **Project/portfolio page URL** — any public URL describing a project
+- **Research paper URL** — arXiv, ACL, IEEE, journal page
+- **Plain text** — freeform description of a project, role, or skill to add
+
+Multiple items can be passed in one call separated by commas or newlines.
+
+## Pipeline
+
+### Step 1 — Classify the input
+
+For each item, determine its type:
+
+| Signal | Classification |
+|--------|----------------|
+| `github.com` URL | **project** (fetch repo info + README) |
+| arXiv / ACL / DOI / journal URL | **publication** |
+| Company name + role + dates in text | **experience** |
+| Tech list (comma-separated tools/languages) | **skill** |
+| Any other URL or rich description | **project** |
+
+### Step 2 — Fetch content (for URLs)
+
+**GitHub repos:**
+1. Use WebFetch to `https://api.github.com/repos/{owner}/{repo}` — extract: `description`, `language`, `topics`, `stargazers_count`, `html_url`
+2. Use WebFetch on the raw README (`https://raw.githubusercontent.com/{owner}/{repo}/main/README.md` or `master`) — extract tech stack, features, key outcomes
+3. If both fail, use WebFetch on the HTML page directly
+
+**Other URLs:**
+1. Try `browser_navigate` + `browser_snapshot` (Playwright) to render JS-heavy pages
+2. Fall back to WebFetch for static pages
+
+**Never invent content.** Only use what is demonstrably on the page. If a page is inaccessible, ask the user to paste the description manually.
+
+### Step 3 — Extract structured data
+
+Extract as much as available. Mark missing fields as `null`:
+
+```
+name:         (string) project/company/paper name
+type:         project | experience | publication | skill
+url:          (string|null) canonical URL
+dates:        (string|null) e.g. "Jan 2025 – Mar 2025" or "2024"
+tech_stack:   (string[]|null) languages, frameworks, tools
+description:  (string) 1-2 sentence summary
+bullets:      (string[]) 2-4 achievement bullets — quantified where possible
+hero_metric:  (string|null) the single best proof point (e.g. "84% accuracy", "2K stars")
+```
+
+**For publications:** also extract `venue` (conference/journal name) and `co-authors`.
+**For experience:** also extract `company`, `location`, `role_title`.
+
+### Step 4 — Generate bullets
+
+Write 2-4 bullets using the same style as the existing `cv.md`:
+- Action verb first, past tense (present tense only for current roles)
+- Quantify outcomes where the source provides numbers
+- NEVER invent metrics — use what is on the page or leave unquantified
+- Keep it concise and ATS-friendly (no buzzwords)
+
+Example transformation:
+> README says "used PyTorch to train a model that classifies 10K images per second"
+→ `Trained PyTorch image classifier achieving 10K img/sec throughput`
+
+### Step 5 — Show a preview and ask for confirmation
+
+Before writing anything, display a formatted preview:
+
+```
+── PREVIEW ──────────────────────────────────────
+
+[PROJECT] Semantic Search Engine
+URL: https://github.com/janesmith/semantic-search
+Dates: Jan 2025 – Mar 2025
+Tech: Python, PyTorch, FAISS, FastAPI
+Bullets:
+  • Built semantic search engine over 500K documents using dense retrieval (FAISS + bi-encoder)
+  • Reduced average query latency from 420ms to 38ms via quantized embeddings
+  • Deployed as FastAPI service with 99.9% uptime over 90-day production run
+
+This will be added to:
+  → cv.md (Projects section)
+  → article-digest.md (proof points)
+
+Proceed? [yes / edit / skip]
+─────────────────────────────────────────────────
+```
+
+- **yes / y** → write to files
+- **edit** → ask what to change, then show preview again
+- **skip / no** → skip this item, move to next
+
+### Step 6 — Write to cv.md
+
+Locate the correct section by type:
+
+| Type | Section in cv.md |
+|------|-----------------|
+| `project` | `## Projects` |
+| `experience` | `## Experience` |
+| `publication` | `## Experience` (or `## Publications` if it exists) |
+| `skill` | `## Technical Skills` (merge into existing categories) |
+| `education` | `## Education` |
+
+**NEVER duplicate.** Before adding, check if the project/company/role name already appears in cv.md. If it does, ask the user whether to update the existing entry or skip.
+
+**Format for projects** (match existing cv.md style):
+```markdown
+### {name} | {dates}
+**Technologies:** {tech_stack joined by ", "}
+- {bullet 1}
+- {bullet 2}
+- {bullet 3}
+```
+
+**Format for experience:**
+```markdown
+### {role_title} | {dates}
+**{company}**
+- {bullet 1}
+- {bullet 2}
+- {bullet 3}
+```
+
+**Format for publications:**
+```markdown
+### {title} | {year}
+**{venue}** — co-authored with {co-authors}
+- {contribution bullet}
+- URL: {url}
+```
+
+**Format for skills** — merge into the existing `**Category:**` lines. If the tech doesn't fit an existing category, add a new one.
+
+**Placement within section:** Add new projects/experience at the TOP of their section (most recent first). Skills are merged alphabetically within each category.
+
+### Step 7 — Write to article-digest.md
+
+If `article-digest.md` does not exist, create it with this header:
+```markdown
+# Article & Project Digest
+
+Compact proof points for use during CV generation and evaluations.
+```
+
+Append a new entry:
+```markdown
+## {name}
+
+- **Type:** {type}
+- **URL:** {url}
+- **Dates:** {dates}
+- **Tech:** {tech_stack}
+- **Hero metric:** {hero_metric}
+- **Summary:** {description}
+- **Proof points:**
+  - {bullet 1}
+  - {bullet 2}
+  - {bullet 3}
+```
+
+### Step 8 — Confirm and summarize
+
+After writing, report:
+
+```
+✅ Added to cv.md:
+   → Projects: Semantic Search Engine (Jan 2025 – Mar 2025)
+
+✅ Added to article-digest.md:
+   → Semantic Search Engine — hero metric: "38ms query latency"
+
+Future CVs will now include this project and prioritize it for relevant roles.
+```
+
+---
+
+## Rules
+
+- **NEVER invent data** that is not on the source page or provided by the user
+- **NEVER overwrite** existing entries — only add or merge
+- **ALWAYS confirm** before writing (Step 5)
+- **Data goes to cv.md and article-digest.md** — these are user-layer files, correct by design
+- If the URL is inaccessible, ask the user to paste the content instead of guessing
+- If dates are unknown, omit them rather than guessing
+- Run `node cv-sync-check.mjs` silently after writing and report any warnings

modes/auto-pipeline.md

@@ -26,8 +26,8 @@ Include Block G in the saved report. Add `**Legitimacy:** {tier}` to the report
 ## Paso 3 — Generar PDF
 Read `config/profile.yml`. Check `cv.output_format`:
 
-- If `"latex"`, execute the full pipeline from `modes/latex.md`
-- Otherwise (default), execute the full pipeline from `modes/pdf.md`
+- If `"html"`, execute the full pipeline from `modes/pdf.md`
+- Otherwise (default), execute the full pipeline from `modes/latex.md` (Jake's Resume template)
 
 ## Paso 4 — Draft Application Answers (solo si score >= 4.5)
 

modes/latex.md

@@ -28,7 +28,6 @@ The template at `templates/cv-template.tex` uses `{{PLACEHOLDER}}` syntax:
 | Placeholder | Source |
 |-------------|--------|
 | `{{NAME}}` | `profile.yml → candidate.full_name` |
-| `{{CONTACT_LINE}}` | Phone / City, State / Visa status — built from profile.yml |
 | `{{EMAIL_URL}}` | Raw email for `mailto:` URL — must not be LaTeX-escaped (from profile.yml) |
 | `{{EMAIL_DISPLAY}}` | Escaped email for display text — LaTeX-special chars like `_` must be escaped, e.g. `first\_name@example.com` |
 | `{{LINKEDIN_URL}}` | Full URL with scheme for `\href{}`: e.g. `https://linkedin.com/in/username`. If `profile.yml` stores a bare host+path (no scheme), prepend `https://` before substitution. |
@@ -123,7 +122,7 @@ All text content MUST be escaped for LaTeX before insertion:
 ## ATS Rules (same as pdf mode)
 
 - Single-column layout (enforced by template)
-- Standard section headers: Education, Work Experience, Personal Projects, Technical Skills
+- Standard section headers: Education, Experience, Projects, Technical Skills
 - UTF-8, machine-readable via `\pdfgentounicode=1`
 - Keywords distributed: first bullet of each role, skills section
 - No images, no graphics, no color in body text
@@ -142,6 +141,6 @@ Same ethical rules as `modes/pdf.md`:
 The generated `.tex` file uses only standard CTAN packages (no custom or bundled dependencies):
 
 - `latexsym`, `fullpage`, `titlesec`, `marvosym`, `color`, `verbatim`, `enumitem`
-- `hyperref`, `fancyhdr`, `babel`, `tabularx`, `fontawesome5`, `multicol`, `glyphtounicode`
+- `hyperref`, `fancyhdr`, `babel`, `tabularx`, `glyphtounicode`
 
 Upload the `.tex` file directly to Overleaf — compiles with no extra configuration.