benchmark

Run each collector script against its fixture, measure time and output size, detect regressions. Trigger: /benchmark

SKILL.md

GitHub

/benchmark

Performance and correctness regression detection for collector scripts.

PROTECTION RULE - never ~/.claude/

Read-only skill. Guard required if write mode added later.

What this is for

  • Scripts that error on their own fixture
  • Output size anomalies (empty or bloated JSON)
  • Execution time regressions
  • Read-only skill. No code changes.

What You Must Do When Invoked

Step 1

-ProjectDir target

Step 2

Run collector: execute each script against its fixture, capture time + output + exit code

Step 3

Report failures, anomalies, slow scripts

Step 4

Write benchmark-report.md

Usage

/benchmark                # interactive
/benchmark <dir>          # scan project
/benchmark -help

During analysis, assign a confidence level to each finding: proven (confirmed by evidence), likely (strong signal, needs review), or suspected (weak signal).