You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
v1.6.0 adds a local-first Skill Usage analytics system to Skill Flow.
The release identifies attributable Skill invocations from supported Agent session, database, and trace formats, stores normalized observations locally, and presents them through a new interactive macOS Usage dashboard.
Highlights
1. Cross-Agent Skill Usage collection
Adds modular collectors for supported Codex, Claude Code, ZCode, OpenCode, Pi, Kimi Code, and WorkBuddy evidence sources.
Uses Agent-specific parsing rules instead of treating generic tool calls, prompts, task lists, or MCP calls as Skill usage.
Includes active and archived Codex sessions and preserves parser coverage diagnostics for Agents without a safe primary counting source.
2. Explainable local observations
Normalizes accepted calls into local observations with Skill, Agent, timestamp, project, session, evidence kind, and source identity.
Counts every accepted Skill invocation in the Skill run total; unique Skill metrics remain separately de-duplicated.
Extracts only whitelisted metadata required for attribution and does not import prompt, response, tool output, or transcript bodies into Usage storage.
3. Interactive desktop Usage dashboard
Adds Today, 24H, 7D, 30D, 90D, and custom time ranges.
Adds fixed-square hourly activity, daily trend bands, hover details, total metrics, top Skills, and top Agents.
Skill and Agent selections update the trend chart, rankings, and per-Agent/per-Skill counts from the same filtered snapshot.
4. Shared analytics contract
Adds Usage storage, analytics services, bridge commands, payload decoding, and desktop view data around one snapshot contract.
Keeps time filtering, Skill identity, Agent aggregation, project attribution, and chart totals consistent across runtime and UI consumers.
User-visible changes
Users can open Usage from the macOS home toolbar and inspect how often Skills run, when they run, which projects they belong to, and which Agents invoked them.
Users can compare overall activity or focus the dashboard on one Skill or Agent without exporting local conversation data.
Usage totals reconcile across the chart, summary metrics, Skill ranking, and Agent ranking for the selected time range.
