JD Filter - Job Description Analysis Tool

Quickly evaluate job descriptions against user-defined filter criteria and output structured summaries.

Workflow

1. Check Filter at Start of Conversation

First use or when explicitly requested:

• •Read user_filter.json to get user's filter criteria
• •If filter not set, prompt user to configure it first

Daily use:

• •Auto-read user_filter.json before analyzing JD
• •No need to confirm with user each time, apply saved criteria directly

2. Auto-Detect and Update Filter

Trigger conditions: When user mentions new filter requirements in conversation. Trigger keywords include:

• •"My requirements are...", "I want...", "I don't want..."
• •"Filter for...", "Only look at...", "Exclude..."
• •"Salary must be...", "Location must be...", "Cannot have..."
• •"Update my filter", "Modify filter criteria", "Save my preferences"

Update process:

1. •Read current user_filter.json
2. •Parse new requirements from user
3. •Merge into existing filter (new requirements take priority)
4. •Update user_filter.json file
5. •Briefly confirm: "✅ Filter criteria updated and saved"
6. •Then continue processing JD analysis (if user provided JD simultaneously)

Common filter categories:

• •location: Location requirements (must_include, must_exclude)
• •requirements: Hard requirements (certifications, student_status, experience, etc.)
• •application_philosophy: Application strategy (approach, threshold, avoid list)
• •user_background: User background (education, skills, projects)
• •ideal_positions: List of ideal position types
• •salary_range: Salary range
• •work_authorization: Visa/work permit requirements
• •company_preferences: Company preferences (size, industry, exclude list)

Example: User says: "I don't want to apply to positions paying less than $25/hour" → Auto-update salary_range.minimum to 25 in user_filter.json → Reply: "✅ Filter criteria updated and saved: Minimum salary requirement $25/hour" → If user provided JD simultaneously, continue with analysis

3. Analyze JD

When user pastes a JD, process with following steps:

Step 1: Extract Key Information

Extract from JD:

• •Location (city, remote policy)
• •Job Scope (core responsibilities, 2-3 sentence summary)
• •Requirements (must-have vs nice-to-have, education, experience, skills)
• •Visa/Sponsorship information (if mentioned)
• •Other filter-relevant information

Step 2: Compare Against Filter

Compare extracted information against user's filter criteria point by point.

Step 3: Output Results

Output format (REJECT):

## 🔴 REJECT

**Rejection Reasons:**
- [Core reason 1: Field mismatch/Skill waste/Career regression, etc.]
- [Core reason 2 (if applicable)]

**📍 Location:** [Location] (✅/❌)
**💼 Job Scope:** [1-2 sentence core responsibilities]
**📋 Requirements:** [3-5 key requirements, comma-separated]

Output format (PASS):

## 🟢 PASS

**📍 Location:** [Location] (✅)
**💼 Job Scope:** [1-2 sentence core responsibilities]
**📋 Requirements:** [3-5 key requirements, comma-separated]

**Match Score:** [Why suitable, 1-2 sentences]
**Application Tip:** [Cover letter key points, 1 sentence]

Filtering Philosophy

User's filtering strategy:

• •Threshold: minimum_requirements_met - Apply if minimum requirements are met
• •Don't over-filter: If user meets minimum qualifications, should PASS
• •Avoid subjective judgment: Don't judge position "quality", "whether it's real data analysis", etc.
• •Example: If JD requires "Bachelor's in Statistics, SQL, Excel" and user has Master's in Statistics, knows SQL and Excel, even if position is mainly reporting rather than advanced analytics, should still PASS

Common over-filtering mistakes:

• •❌ "This position seems more like reporting than real data analysis" → PASS if user meets requirements
• •❌ "Meets requirements but position quality is low" → Not our scope to judge
• •❌ "Position tech stack not cutting-edge enough" → Only check filter compliance, don't evaluate position value

Important Notes

• •Base judgment on explicitly stated JD information, don't speculate
• •If JD doesn't mention filter-relevant info (e.g., sponsorship not mentioned), mark as "Not mentioned"
• •Keep summaries concise, each section within 2-3 lines
• •Output in English, but preserve proper nouns from JD (position titles, company names, tech stack)