monitor-ai-tooling
Monitor AI tooling
Run an on-demand report from existing measurements. Collection stays deterministic; this skill does not install tools or start recurring model analysis.
Process
1. Establish the evidence
Read the project's AI tooling record when AGENTS.md points to one, the requested project and period, and the available local measurement sources.
Use configured tool analytics, saved verification results, or an export from the user's existing monitor.
Inspect source schemas and installed help before choosing read-only export commands; preserve source files and redact credentials or unrelated session content from the report.
For an output-filtering tool (RTK's rtk gain, for example), inspect its savings and history or export options without changing counters or configuration.
Prefer local collection without network requests or additional model calls.
If setup or data is missing, produce a gaps report: name the gaps a future measurement period fills, and the setup each missing source needs.
Treat no observations as unknown usage, not proof a tool was unused.
Done when every source has a provenance, unit, period, scope, and project attribution or an explicit attribution gap.