Skill UI

Compares prompt changes against a fixed evaluation set, recording model settings, checks, missing results, judge limitations, and case-level regression evidence for a reviewable decision.

Designs a minimal performance review system for SMEs, capturing cycle, content, and process. Generates CSV, SQL, JSON Schema, or Notion artifacts on request, referencing KPI and OKR data. Ideal for appraisals, review cycles, and employee evaluation tracking.

A public-safe preliminary screen for challenges to the impartiality of an AI evaluator, LLM-as-a-judge, or decision-support agent based on prior outputs, suspected bias, unstable reasoning, or self-reinforcing memory. Classifies evidence and recommends proportionate remedies.

Diagnose why an alert rule in Nightingale (n9e) did not fire as expected. Trace the data flow from rule sync to notification, using evaluation logs and query verification to pinpoint the blocking step. For Release 22 and above.

This tool validates AI agent changes by running statistical regression tests to ensure performance doesn't degrade. It grades tool trajectories, detects flaky cases, and integrates with CI to block merges when significant regressions occur. Supports pass@k metrics and paired statistics for reliable evaluation.

This skill executes a single iteration of an automated research experiment. It reviews historical results, decides on a specific code modification, edits the target file, commits changes via git, and evaluates the outcome. Users can specify an experiment or select from a list. The process ensures only one change is made per iteration to isolate variables, preventing evaluation script modification. It supports strategy escalation based on run count and reports keep, discard, or crash status.