create-dashboard

name	create-dashboard
description	Generate an agent-specific `/update-dashboard` skill that keeps `dashboard.yaml` current for Trinity. Analyzes the agent's purpose and data sources, proposes metrics, gets user approval, then scaffolds a schedulable skill.
disable-model-invocation	false
user-invocable	true
allowed-tools	Read, Write, Glob, Bash, AskUserQuestion
metadata	{"version":"1.2","created":"2026-05-27T00:00:00.000Z","author":"Ability.ai","changelog":["1.2: Chart widget templates removed — `type: chart` never existed in Trinity (the agent-server validator strips it, the frontend has no renderer); trend lines come from the platform's Dynamic Dashboards enrichment instead: metric/progress widgets get auto-captured history + sparklines, keyed by a stable `id:` field (now taught). Added markdown widget template and a no-YAML-anchors caution (hardened loader rejects aliases, trinity#1965)","1.1.3: Report guard also swallows the `requires an agent-scoped API key` refusal (Trinity mcp-server reports.ts, in v0.9.0) — a user/admin-key session sees mcp__trinity__report but cannot publish; skip silently, never retry","1.1.2: The generated /update-dashboard is built to run on cron, so it ships disable-model-invocation: false — true made it unreachable to the scheduler. Scheduling instructions replaced: /trinity-schedules is retired, so declare the cron in template.yaml schedules: and reconcile, with the ent#89 literal-true rule and the autonomy gate both called out","1.1.1: Note that reports are a rolling history — pruned past agent_reports_retention_days (default 90 days), not a permanent archive","1.1: Generated /update-dashboard now also emits a guarded {agent}.kpi_snapshot report (display_hint kpi) after writing the dashboard — the same headline numbers accumulate as an append-only history on the Reports tab alongside the live snapshot; skipped silently off-Trinity","1.0: Initial version — generate an agent-specific /update-dashboard skill that gathers metrics from the agent's data sources and writes a schedulable dashboard.yaml for Trinity"]}

/trinity:create-dashboard

ℹ️ First, set expectations: before anything else, print one short line with this skill's version and its most recent change — the top entry of metadata.changelog above — e.g. create-dashboard vX.Y — recent: <summary>. Then proceed.

Generate an agent-specific /update-dashboard skill that keeps dashboard.yaml current for Trinity. Analyzes the agent's purpose and data sources, proposes metrics, gets user approval, then creates a schedulable skill.

Trigger

User wants to:

• Add a dashboard to an existing agent
• Create or regenerate dashboard metrics
• "create dashboard", "add dashboard", "setup dashboard"

What This Creates

A new skill at .claude/skills/update-dashboard/SKILL.md that:

• Gathers current metrics from agent data sources
• Writes dashboard.yaml to /home/developer/dashboard.yaml
• Is designed to run on a schedule (e.g., hourly via Trinity cron)
• Uses widget types appropriate for the agent's purpose

---

PHASE 1: Gather Context

1.1 Read Agent Identity

Read CLAUDE.md (or README.md if no CLAUDE.md exists).

Extract:

• Agent name and purpose
• Primary responsibilities
• Key workflows and capabilities

1.2 Discover Data Sources

Glob for potential data files:

• *.json, *.yaml, *.yml in workspace root
• memory/, data/, logs/, state/ directories
• Any *_log.md, *_state.*, *_history.* files

1.3 Inventory Existing Skills

ls -la .claude/skills/*/SKILL.md 2>/dev/null

Note skill names - they indicate what the agent does.

1.4 Check for Existing Dashboard

Read dashboard.yaml if it exists - use current structure as baseline.

---

PHASE 2: Propose Dashboard Metrics

Based on analysis, propose a dashboard structure. Consider these categories: