take-home-evaluator

What problem does it solve?

Take-home assessments are a core part of engineering hiring, but inconsistent, unanchored evaluation leads to biased panel debriefs, generic feedback, and unfair candidate outcomes. This Skill eliminates that by tying every score to rubric-defined evidence and keeping hire decisions with the human panel.

Core Features & Use Cases

• Rubric-anchored scoring: Every dimension score is backed by verbatim citations from the candidate's submission (file path, line range, exact content) so feedback is evidence-based, not subjective.
• Deterministic pre-checks: Runs build, test, and lint commands first to separate auditable, objective results from LLM judgment, with sandboxing guardrails for unsafe candidate code.
• Policy-calibrated AI-use signals: Detects potential AI-use patterns calibrated to the candidate's disclosed policy, surfacing them as discussion notes for the panel rather than automatic verdicts.
• Use Case: An engineering hiring panel can use this Skill to generate consistent, fair evaluation reports for multiple take-home submissions per role, reducing evaluation bias and ensuring all panelists work from the same rubric anchors.

Quick Start

Use the take-home-evaluator skill to score a candidate's take-home submission against the provided role rubric and generate a structured, evidence-backed evaluation report for your panel debrief.