Candidate Screening Rubric

• Outcome: Pass
• Latest audit: May 28, 2026, 10:57 AM
• Version:
  0.1.2

Overview

This is a markdown-only recruiting screening aid that asks for candidate materials, applies evidence-based scoring and bias checks, and does not install code or take actions on its own.

Use this only with appropriate hiring-process controls. Redact names, emails, phone numbers, exact addresses, photos, and other direct identifiers where possible; verify that outputs cite supplied evidence; and treat the rubric as a draft aid for structured human review, not an automated hiring decision.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (1)

Autonomous Decision Making

Medium

• Category: Excessive Agency

Content

| --- | --- |
| Affinity / similarity | Did the score lean on shared school, employer, or hometown? Remove and rescore on evidence. |
| Halo / horn | Are all items uniformly strong or uniformly weak without differentiation? Add the contrary evidence the materials contain. |
| Recency / tenure heuristic | Did the score penalize a gap or a short tenure without checking the supplied explanation? Re-anchor to evidence the candidate or call notes provided. |
| Name / photo / demographic | Did any inference draw on a name, photo, gender, age signal, accent, parental status, disability, or other protected characteristic? Remove without exception. |
| Pattern-matching to a "typical" hire | Did the rubric implicitly require the candidate to look like prior hires (same title path, same companies)? Restate the requirement in skills/outcomes terms and rescore. |
| Generic praise / generic criticism | Does any rationale use vague language ("smart", "not technical enough") without citing evidence? Replace with the cited line. |

• Confidence: 75% confidence
• Finding: without checking

VirusTotal

59/59 vendors flagged this skill as clean.