insightslisted

# sessions:insights You are the **conductor** over the local analytics engines. There is no separate `/trends` or `/perf` plugin command — mix lives under `agents insights mix`. This skill decides which engines to run, stitches their output, and returns a short list of **evidence-backed actions**. Never upload raw transcripts. Everything below stays on-machine unless the user explicitly asked for a shareable HTML artifact. ## Engines (orchestrate; do not invent a parallel pipeline) | Engine | CLI | Answers | |---|---|---| | **How you work** | `agents insights` | Tools, friction, rhythm; default group-by account (Claude attribution; other harnesses under `unattributed:<agent>`) | | **Usage mix** | `agents insights mix` | Harness/model mix, token ratios, session volume, secrets-hot, browser activity | | **Latency / friction** | `agents insights perf` | Slow hooks/commands/runs; `agents insights perf friction` for guard-block loops | | **Resource dead weight** | `agents sessions stats` | Explicit skill/command invocations; installed-but-never-invoked | Related (only if the ask needs them; not the default path): - `agents cost` — what it cost - `agents output` — what shipped - `agents view` — live quota If `$ARGUMENTS` already names one engine (`mix`, `perf`, `stats`, `insights` alone), run that engine and still synthesize actions — do not dump a raw table as the whole answer. ## Default recipe Forward user flags (`--since`, `--agent`, `--project`, `--json`, `--narrative