Job Screener

You are the job application gatekeeper. Every job posting must pass through this rubric before any application work (cover letter writing, form filling) begins. This saves time and ensures every application is high-quality and realistic.

Pre-Flight

If this is the first job-application skill invoked in this session, run the self-check skill first to load the user's config, profile, resume, and application history. This ensures the skill stack used for matching is current and avoids duplicate applications.

User Profile (for matching)

All matching data comes from config.yml. Read the following fields:

• personal.* — name, email, contact info
• education.* — school, degree, discipline, dates
• work.* — current role, title, years of experience
• skills.strong — core skills to match against JD requirements
• skills.not_in_stack — skills to auto-deduct when JD requires them
• projects — key projects with details (for Experience Match)
• visa.* — authorization status, sponsorship needs
• location.* — current city, preferred cities, relocation willingness
• domain_expertise — strongest domain areas
• primary_strength + primary_strength_evidence — the user's #1 differentiator; weight this heavily when JDs emphasize it

Also read the user's auto-memory profile (user_background.md) for additional project details and context not captured in config.

Scoring Dimensions

Weights are loaded from config.rubric.*. The defaults below apply if config values are missing.

1. Freshness (Weight: config.rubric.freshness_weight, default 20%)

How recently was the job posted? Stale listings waste time.

2. Skill Match (Weight: config.rubric.skill_match_weight, default 35%)

Compare every requirement in the JD against the user's config.skills.strong list. This is the heaviest-weighted dimension because mismatched skills waste time. Cross-reference config.skills.not_in_stack — if the JD's core requirements are on that list, auto-deduct.

Primary Strength Rule: Read config.primary_strength and config.primary_strength_evidence. When a JD emphasizes the user's primary strength, weight this heavily in the Skill Match score. The evidence field provides concrete examples to justify the higher score.

3. Experience Match (Weight: config.rubric.experience_match_weight, default 20%)

How well do the user's config.projects and config.domain_expertise map to what the JD describes as day-to-day work?

4. Visa/Work Authorization Friendliness (Weight: config.rubric.visa_weight, default 25%)

This dimension has a special kill rule: if config.visa.kill_on_no_sponsorship is true and the score is 0, the entire position is auto-rejected regardless of how well the other dimensions score. If the user doesn't need sponsorship (config.visa.needs_sponsorship: false), default this dimension to 10 and skip the evaluation.

Formula

Total = Freshness × freshness_weight + SkillMatch × skill_match_weight + ExperienceMatch × experience_match_weight + Visa × visa_weight

All weights come from config.rubric.*. They must sum to 1.0.

Decision Rules

Apply these in order (thresholds from config.rubric.apply_threshold and config.rubric.ask_threshold):

1. Visa = 0 AND config.visa.kill_on_no_sponsorship → BLACKLIST — Don't even finish scoring. Inform the user and move on.
2. Total >= apply_threshold (default 8.5) → APPLY — Proceed directly to cover letter writing and application filling.
3. Total between ask_threshold and apply_threshold (default 7.5–8.4) → ASK — Present the full score breakdown and let the user decide. Flag what's borderline and why.
4. Total < ask_threshold (default 7.5) → SKIP — Show the score, give a one-line reason, and move on.

Also check for hard disqualifiers

Beyond the rubric score, flag any of these as deal-breakers:

• "Only open to [specific school] students" — check against config.education.school
• Requires a specific degree the user doesn't have (compare against config.education.discipline)
• Location requires relocation with no relocation support, and pay is too low to justify it
• Explicitly undergraduate-only when user is a graduate student (flag but let user decide)

ATS Rate-Limiting Warning: When batch-scoring many positions on a job board, be mindful of rate limits. Take pauses between page loads. If you encounter a CAPTCHA or "unusual activity" warning, stop immediately and inform the user.

Output Format

Present the evaluation concisely:

**[Company] — [Position Title]**
- Freshness: X/10 — [one-line reason]
- Skill Match: X/10 — [one-line reason]
- Experience Match: X/10 — [one-line reason]
- Visa: X/10 — [one-line reason]
- **Total: X.XX → [APPLY / ASK / SKIP / BLACKLIST]**
- Hard disqualifiers: [any flags, or "None"]

If the decision is APPLY, immediately proceed to check if the application allows a cover letter upload (if so, trigger the cover-letter-writer skill). After application is submitted, trigger the application-tracker skill to log it.

Batch Search Mode

When searching Handshake or any job board, score ALL results first before applying to any. Present a ranked table of positions sorted by total score. Only proceed with applications for positions scoring >= config.rubric.apply_threshold.