Installs into .claude/skills of the current project.
Are you the author of Threshold Sweep?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/yogsoth-ai-threshold-sweep)
---
name: threshold-sweep
description: Compute consensus status at multiple threshold levels to produce a threshold-consensus curve.
execution: subagent
prompt: ./prompt.md
input: judgments[], threshold_range
used-by: structured-consensus
---
# Threshold Sweep
Systematically compute consensus status at multiple threshold levels across a specified range. Produces a curve showing how many items achieve consensus as the threshold varies, revealing robust vs. fragile consensus items.
## Execution
Spawn a subagent that takes judgments and a threshold range, then computes consensus classification at each threshold level to produce the curve.
## Why Subagent
- Sweep computation is a bounded, parallelizable task
- Produces a standardized curve structure
- Keeps orchestration context clean
## HARD-GATE
Output MUST contain: `threshold_curve` with at least 5 data points, each showing threshold value and number of consensus items. Curve must span the full input range.
