Heuristic Analysis vs. Rigid Filtering: PRAETOR v5.5 🛡️
Resume screening is often criticized for being a "black box." Many automated systems rely on rigid keyword matching, which can inadvertently penalize candidates with non-linear career paths or those who prioritize privacy.
To explore an alternative, I developed PRAETOR v5.5, a free and experimental system prompt designed to test heuristic evaluation logic for CV self-assessment.
The Objective: Neutralizing the "Gap" Penalty
It is well-documented that some recruitment algorithms slash scores for career gaps exceeding 6 months. This often overlooks the human context—such as caregiving, health issues, or further education.
PRAETOR is a technical experiment in bias-aware prompting. It is designed to evaluate resume alignment without defaulting to penalties for structural gaps, provided context is available.
Project Overview: PRAETOR v5.5 (Experimental)
This is a 100% free, open-source System Prompt intended for personal reflection. It is not a software product, but a set of heuristic instructions for LLMs (like GPT-4 or Claude).
Implementation Details:
1. ⚖️ Heuristic Experience Scoring
The prompt includes a specific "No Penalty" rule. When a user provides context for a career gap (e.g., maternity leave or health-related hiatus), the logic is instructed to neutralize any score deduction. The focus shifts to the relevance of the experience rather than its mere continuity.
2. 🛡️ Privacy & PII Detection Protocol
Data hygiene is a concern when using public LLMs. This prompt includes a safety module designed to detect Personal Identifiable Information (PII) like phone numbers and physical addresses. It warns the user to redact such data before the analysis begins.
3. 🧠 Tiered Semantic Coverage
Rather than a simple keyword count, PRAETOR uses a tiered model to match skills based on semantic meaning. For example, it is instructed to recognize that "ML" and "Machine Learning" represent the same competency, preventing score loss due to formatting differences.
Why Open Source?
I am sharing this logic to encourage transparency in how AI handles career data. Being a system prompt, the scoring weights and bias-mitigation rules are fully auditable by anyone.
🔗 Access the Prompt
As an experimental tool, PRAETOR requires peer review. If you are interested in Prompt Engineering or the ethics of AI in recruitment, I would value your feedback on the heuristic weights.
👉 GitHub Repository: https://github.com/simonesan-afk/CV-Praetorian-Guard
Disclaimer: This tool is for educational self-reflection and personal use only. It is explicitly designed NOT to be used for hiring, screening, or any professional employment-related decisions.
Top comments (0)