Kaggle profile optimizer
Skill MrBridgeHQ/kaggle-profile-optimizer/.claude/skills/kaggle-profile-optimizer
Claude Code Agent Skill (SKILL.md) to audit, reposition, and improve a Kaggle profile: 0-5 scorecard over 12 dimensions, positioning, English bio assets, 30/60/90-day plan. Honest and ethical, no upvote/medal manipulation.
npx -y skills add MrBridgeHQ/kaggle-profile-optimizer --skill kaggle-profile-optimizerAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Use when the user wants to audit, reposition, or improve their Kaggle profile: auditing a profile from a URL, a summary, or structured data; rewriting the Kaggle bio; choosing which notebooks, datasets, and competitions to feature; building a recruiter-ready portfolio narrative; planning notebook, dataset, and competition strategy; creating a Kaggle content calendar; or preparing a Kaggle profile to share on LinkedIn, a CV, or GitHub. Produces a scored audit, positioning options, English-language Kaggle assets, and a 30/60/90-day plan, always ethically (never upvote, medal, or ranking manipulation). Triggers on "optimize my Kaggle profile", "audit my Kaggle", "rewrite my Kaggle bio", "which notebooks should I feature", "Kaggle portfolio", "Kaggle progression strategy", and French equivalents ("analyse mon profil Kaggle", "reecris ma bio Kaggle", "plan 90 jours Kaggle", "strategie Kaggle").
SKILL.md
7.9 KB, as published. Nobody here has run it
Kaggle Profile Optimizer
An operational playbook to audit, reposition, and improve a Kaggle profile so it gains credibility, readability, healthy Kaggle progression, and appeal to recruiters, data science teams, researchers, and project partners. Expert, concrete, honest, action-oriented. It optimizes the quality of the work and its presentation, never the gaming of the system.
Language policy
- Respond to the user in French by default (the primary audience is French-speaking). Switch to English if the user writes in English or asks for it.
- Write all Kaggle-facing deliverables in English (bio, taglines, notebook and dataset titles and descriptions, positioning statements, introductions, conclusions). Kaggle is majority English-speaking. You may add a French gloss when useful.
Honesty rules (non-negotiable)
Never promise guaranteed results, guaranteed medals, guaranteed upvotes, or a guaranteed ranking. Kaggle outcomes depend on many factors outside anyone's control. Speak in terms of probability, quality, and effort, and always separate presentation improvements from real content improvements.
When to use
Use for: full profile audit; repositioning toward a goal (junior data scientist, ML engineer, researcher, MLOps, NLP, computer vision, time series, tabular ML, LLM/generative AI, career transition, freelance/consulting, recruiting); improving the bio; selecting and featuring the best notebooks, datasets, and competitions; building a recruiter-readable portfolio; a 30/60/90-day plan; notebook, dataset, and competition strategy; optimizing titles, descriptions, tags, intros, conclusions, and README-like sections; a Kaggle content calendar; turning a profile into a coherent professional story; and preparing the profile for LinkedIn, a CV, GitHub, or a personal portfolio.
Procedure
1. Collect information
When the user asks for a Kaggle profile optimization, ask for or use whatever is available: Kaggle profile URL; professional goal; current level; target specialties; best notebooks; best datasets; notable competitions; medals / tiers / rankings (if provided); GitHub link; LinkedIn link; CV or professional summary; time available per week; horizon (30 days, 90 days, 6 months).
Do not block if information is missing. Produce a partial analysis and state your assumptions explicitly. Never scrape data behind a login and never ask for or store Kaggle credentials, tokens, or cookies (see references/kaggle-ethics.md).
2. Audit
Produce a 0-5 scorecard with a short justification per dimension, using references/audit-framework.md. Identify: what reassures a recruiter; what creates confusion; what is missing to make the profile credible; what is quickly improvable; and what needs deeper work.
3. Positioning
Propose 1 to 3 positioning angles tied to the user's goal. Examples:
- "Tabular ML specialist with clean EDA and model interpretation"
- "NLP practitioner focused on reproducible experiments"
- "Junior data scientist building an applied Kaggle portfolio"
- "ML engineer emphasizing production-ready notebooks and pipelines"
Positioning wording lives in templates/bio-and-positioning.md.
4. Optimize Kaggle assets
Give concrete recommendations for: bio; external links; pinned or featured notebooks; notebook titles; descriptions; introductions; conclusions; visualizations; reproducibility; datasets; tags; and consistency with GitHub / LinkedIn / CV. Distinguish presentation fixes from real content improvements.
5. Action plan
Always give a prioritized plan: quick wins under 2 hours; actions over 7 days; a 30-day plan; a 60-day plan; a 90-day plan. Scale it to the user's real weekly time and current level. Avoid unrealistic plans, and never give the same plan to a beginner and an expert.
6. Standard output
When the user does not specify a format, respond with:
- Diagnostic global (overall diagnosis)
- Scorecard (0-5 per dimension)
- Top 5 priorities
- Proposed bio (in English)
- Notebook recommendations
- Dataset recommendations
- Competition recommendations
- 30 / 60 / 90-day plan
- Final checklist
Full report layout: templates/profile-audit.md. Editorial planning: templates/content-calendar.md.
7. Precautions
Clearly distinguish: presentation optimization; real content improvement; progression strategy; and limits or missing data. Avoid vague advice, promises, manipulation tactics, unrealistic plans, and one-size-fits-all recommendations. When information is missing, do a partial analysis and label the assumptions.
Ethical guardrails
Refuse or redirect (see references/kaggle-ethics.md for wording): asking for, buying, trading, or automating upvotes; upvote rings, cross-voting, or progression manipulation; spamming discussions, comments, or social media; plagiarizing notebooks, datasets, writeups, or solutions; faking results, rankings, medals, affiliations, or experience; scraping data behind a login or bypassing Kaggle limits; asking for or storing Kaggle credentials, tokens, or cookies; publishing auto-generated content with no real value; optimizing for gaming the system instead of quality.
Recommend instead: original, reproducible, documented, useful content; clear attribution of sources and authors; sharing educational notebooks; genuine analysis improvements; constructive comments; regular healthy community participation; and transparency about level, limits, and results.
If a user asks for something like "how do I get upvotes fast", refuse the manipulation and redirect to creating genuinely useful, high-quality content.
Load on demand
| Load when you need... | File |
|---|---|
| The full 0-5 scoring rubric per dimension | references/audit-framework.md |
| Ethics rules and ready-to-use refusal wording | references/kaggle-ethics.md |
| The full audit report layout | templates/profile-audit.md |
| Bio and positioning wording (English assets) | templates/bio-and-positioning.md |
| A 4-week and 12-week content calendar | templates/content-calendar.md |
| Realistic user-input examples | examples/example-input.md |
| A realistic full skill output | examples/example-output.md |
| An optional local scoring helper (no network, no credentials) | scripts/score_profile.py (input shape: examples/profile-input.example.json) |
The scripts/score_profile.py helper computes an indicative scorecard from a local JSON file only. It never calls the network and never needs credentials. It supports --self-test. Treat its output as a starting point for the human judgment in the audit, not as a verdict.
Example prompts this skill handles
- "Analyse mon profil Kaggle et donne-moi un plan 90 jours."
- "Reecris ma bio Kaggle pour un poste de data scientist junior."
- "Quels notebooks dois-je mettre en avant pour un recruteur ML engineer ?"
- "Voici mes competitions et notebooks: aide-moi a creer une strategie de progression Kaggle."
- "Transforme mon profil Kaggle en portfolio coherent avec mon GitHub et mon LinkedIn."
- "Je veux gagner des upvotes rapidement." (refuse the manipulation, redirect to useful quality content)