Resume Tailoring Skill
Researches each role, surfaces your undocumented experience, and generates tailored, truth-preserving resumes from your existing resume library.
Reads only the local resume library and writes to user directories; no credentials or dangerous operations. Checkpoints at template, matching, and generation stages; depends on document-skills sub-skill; data flow reasonably disclosed. Deducted for: no explicit rollback mechanism, minimal-privilege detail lacking for library-update writes, undeclared boundaries on external data use (WebSearch/LinkedIn scraping), and unverified publisher.
Workflow is self-consistent with graceful degradation (JD-only fallback without WebSearch, markdown-only without plugin) and decent failure feedback. Deducted for: core logic is pseudo-code only, no executable matching implementation; SKILL.md is truncated mid-section, indicating incomplete documentation; static review cannot exercise key paths; sub-skill availability unverified.
Triggers are clear (user provides JD + existing markdown resume library) with explicit non-fit cases (writing from scratch, cover letters, LinkedIn). Deducted for: multi-job detection relies on English phrases, no Chinese-language support declared; heavy dependence on WebSearch and LinkedIn scraping may be unreachable from mainland-China networks, and this limitation is not disclosed.
Well-layered docs (SKILL.md + README + support files + docs/), MIT license, contributing guide, troubleshooting, and test checklists. Deducted for: no version numbers or changelog, unclear maintenance ownership and update path; SKILL.md content truncated; referenced support files not present in this evidence set so cross-reference integrity is unverifiable.
Clear goals, directly usable output formats (MD/DOCX/PDF/report), checkpoint design reduces rework cost. Deducted for: effect claims (92% coverage, time savings) are author-supplied examples without independent verification; static review cannot confirm output quality; value claim unverified by execution.
Test checklists and design doc paths exist; regression tests claimed. Deducted for: example results are not reproducible, no CI workflows or committed test suites providing third-party execution evidence, and truncation prevents verifying cited sections (e.g., SKILL.md lines 1244-1320).
- Static review only; no code was executed. All effect claims (JD coverage, time savings) are author-stated and independently unverified.
- The SKILL.md source file is truncated mid-section; documentation may be incomplete — verify the full repository version before use.
- Company research and role benchmarking depend on WebSearch and LinkedIn scraping, which may be unreachable from mainland-China networks; the skill does not disclose this limitation.
- DOCX/PDF generation requires the separately installed document-skills plugin, an external untested dependency.
- No Chinese-language support is declared; multi-job detection relies on English phrases and may not trigger on Chinese input.
- No versioning or changelog; maintenance and update path unclear.
What does this skill do, and when should you use it?
A Claude Code skill that takes a job description plus your Markdown resume library and produces a tailored resume (MD/DOCX/PDF) with a transparency report. It parses the JD, researches the company and role benchmark, runs a branching conversational interview to surface experiences you never documented, then matches library content to a generated template using confidence scores. Its core principle is truth-preserving optimization: reframe and emphasize, never fabricate. Every key decision goes through a user checkpoint, and each successful resume can be saved back to grow the library.
1) Scans resumes/ for Markdown files, parses roles, bullets, skills and education, auto-tags themes/metrics/keywords, and builds an in-memory experience database; 2) Uses web search to parse the JD, research company culture, and benchmark the role via LinkedIn profiles, synthesizing a 'success profile'; 3) Generates a resume template with title reframing options (e.g. Graduate Researcher → Research Software Engineer) and same-company role consolidation advice, confirmed at a checkpoint; 4) Optionally runs a branching interview to surface undocumented experiences and draft bullets; 5) Matches content per template slot with weighted scoring (direct 40% / transferable 30% / adjacent 20% / impact 10%), presenting gaps under 60% confidence honestly with options; 6) Generates Markdown, DOCX (via document-skills:docx), optional PDF, and a report with coverage metrics and interview-prep recommendations; 7) On approval, saves the new resume into the library and rebuilds the database.
- A job seeker with a dozen historical Markdown resumes who wants a high-fit resume for one specific job description
- Applying to 3-5 similar roles at once (e.g. TPM roles at Microsoft, Google, AWS) via batch mode with one shared discovery session
- An internal transfer candidate (e.g. a Microsoft employee applying to a 1ES Principal PM role) who wants internal experience highlighted with internal terminology
- A career changer (e.g. TPM moving to an ecology PM role) needing truthful title reframing and explicit gap analysis
- Someone with an employment gap (e.g. two years founding a startup) who wants it framed honestly as entrepreneurial experience
- Engineers or researchers with rich but undocumented experience who struggle to articulate it on paper
What are this skill's strengths and limitations?
- Explicit truth-preserving principle: reframe only, never fabricate, with transparent gap reporting — important for job-application integrity
- Checkpoint-heavy design gives full user control at template, matching and generation stages
- Multi-job batch mode shares discovery across jobs, saving ~11% time for 3 jobs and ~27% for 5
- Self-improving: approved resumes feed back into the library, enriching future sessions
- Complete output: resume in three formats plus a report with coverage metrics and interview-prep suggestions
- Clear degradation paths: research failure falls back to JD-only analysis; DOCX failure falls back to Markdown
- Hard dependency on an existing Markdown resume library — cannot write from scratch; a small library (1-2 resumes) sharply limits matching options
- Out of scope for cover letters and LinkedIn optimization, per the source itself
- Deep research depends on WebSearch and LinkedIn; obscure companies or platform unavailability degrade quality
- DOCX/PDF require the document-skills plugin; without it, Markdown-only output
- Headline numbers (92% coverage, time savings) come from the author's own examples with no independent validation
- Testing is a manual checklist; the repo shows no automated test suite
How do you install this skill?
Option 1 (recommended): git clone https://github.com/varunr89/resume-tailoring-skill.git ~/.claude/skills/resume-tailoring, then restart Claude Code and verify with /skills that resume-tailoring appears. Option 2 (manual): mkdir -p ~/.claude/skills/resume-tailoring and place all repository files there. Before use, create a resumes/ directory with at least 1-2 Markdown resumes; install the document-skills plugin if you want DOCX/PDF output.
How do you use this skill?
In Claude Code, trigger it with something like: "I want to apply for [Role] at [Company]. Here's the JD: [paste job description]". The skill runs: library build → company/role research → template (checkpoint) → optional experience discovery → scored matching (checkpoint) → MD/DOCX/PDF/Report generation → optional library update. For batch mode, provide multiple JDs (e.g. "I want to apply for these 3 roles: ..."); the skill detects and offers multi-job mode. Every checkpoint allows revision or rollback; output files follow the {Name}_{Company}_{Role}_Resume.md naming pattern.