Task: Skills Collection Quality Audit

Your team maintains a collection of 12 AI agent skills in skills/. There has been no quality audit in six months. A new engineer wants to establish a repeatable monthly audit process that tracks quality trends over time.

Run a batch audit of the entire collection, compare results against a previous baseline stored in .context/audits/, and produce a summary report identifying which skills need remediation before the next sprint.

Context

• Repository root: current working directory
• Previous baseline: .context/audits/*/2025-10-01/audit.json files exist for 8 of the 12 skills
• Four new skills have no baseline yet
• Grading thresholds: A >=126, B+ >=119, B >=112, C/C+ <112 (blocked from publishing)

Output Specification

Produce:

1. audit-execution.sh — commands used to run the batch audit
2. audit-results.json — full JSON output from the batch run
3. audit-report.md — human-readable table: skill | score | grade | delta vs baseline
4. baseline-comparison.md — trend analysis: improvements, regressions, new skills