Agents Dashboard

- the seven specialists, their work & their wear -

Enter PIN to continue

Headline numbers

The roster (click a row to see details)

Agent	What it does	Runs	Success	Last fired

Recent activity (last 25 invocations)

How this works

Where the data comes from

Each agent appends one JSONL line to ai-shared-brain/logs/agent-activity.jsonl at the end of its run with {ts, agent, outcome, summary}. refresh-agents-dashboard.ps1 denormalizes the log + agent frontmatter into data/agents.json for this page.

Refresh

pwsh ~/.claude/scripts/refresh-agents-dashboard.ps1: Regenerate & push the snapshot
... -NoPush: Local-only refresh

Caveats

Self-reported logging — agents that forget to log don't count. Outcome label is the agent's own judgment, not verified. Mid-session agent calls aren't always observable from the parent. Treat numbers as a floor, not a ceiling.