controlled-perturbation
$
npx mdskill add yogsoth-ai/de-anthropocentric-research-engine/controlled-perturbationSystematically perturb parameters to identify degradation thresholds.
- Identifies performance limits by varying parameters along defined axes.
- Relies on subagent execution and external performance measurement tools.
- Uses specified method, variation axis, and range to guide perturbation steps.
- Returns recorded performance data for analysis and threshold detection.
SKILL.md
.github/skills/controlled-perturbationView on GitHub ↗
--- name: controlled-perturbation description: Systematically vary parameters along defined axes, recording performance at each point to identify degradation thresholds. execution: subagent prompt: ./prompt.md input: method (string), variation_axis (string), range (string) used-by: validity-envelope-mapping --- # Controlled Perturbation Systematically perturb to find degradation thresholds. ## Execution Subagent — spawned via subagent-spawning/spawn-agent. ## Budget One unit = one perturbation curve along one axis.