Skill Executions View

The Skill Executions view (/admin/skill-executions) tracks every invocation of research skills (research, research-scan, research-deep). It provides a list page with summary KPIs and a detail page with execution metrics, metadata, linked plan, and an audit timeline.

List Page

Route: /admin/skill-executions

Summary KPI Cards

Four cards summarize execution health:

Card	Metric	Source
Total Executions	Count of all skill execution records	skillExecution.summary
Completion Rate	Percentage of executions in completed status	skillExecution.summary
Avg Duration	Mean duration of completed executions	skillExecution.summary
Recent Failures (24h)	Count of failed executions in the last 24 hours	skillExecution.summary

Below the cards, a breakdown row shows execution counts by skill name (e.g., research: 12, research-scan: 8).

The summary procedure runs four queries: total count, group-by-status counts, group-by-skill-name counts, average duration of completed executions, and a 24-hour failure count using gte(createdAt, oneDayAgo).

Executions Table

A paginated table (20 rows per page) with two filter dropdowns:

Filters:

• Skill -- research, research-scan, research-deep, or all
• Status -- started, executing, completed, failed, or all

Each row shows:

Column	Content
Skill	Skill name, linked to the detail page (/admin/skill-executions/{id})
Status	Color-coded badge
Plan	Linked plan name (links to /admin/research-plans/{planId}), or dash if unlinked
Started	Relative time (e.g., "5m ago", "2h ago")
Duration	Human-readable duration (ms/s/m)
Session	Truncated Claude Code session ID
Error	Truncated error message in red (if failed)

The skillExecution.list procedure joins skill_execution with research_plan via a left join to include the plan name and status.

Detail Page

Route: /admin/skill-executions/{executionId}

Metrics Cards

Four cards across the top:

Card	Content
Status	Color-coded badge
Duration	Formatted duration with clock icon
Started	Start timestamp
Completed	Completion timestamp (or dash if still running)

Error Card

Shown only when the execution has failed status. Displays the error message in a red-bordered card.

Metadata Card

Displays the execution's metadata JSON in a formatted <pre> block. Shows "No metadata recorded." when empty.

Linked Research Plan Card

Shown when the execution has an associated planId. Displays the plan name, status badge, and research question. Includes a "View Plan" button linking to /admin/research-plans/{planId}.

Session Timeline

An audit log table filtered to the execution's Claude Code session ID. Events come from planAuditLog filtered by both planId and claudeCodeSessionId. Each row shows:

Column	Content
Event	Badge with event type
Step	Truncated step ID
Data	Expandable JSON details
Time	Timestamp

Session ID Card

Shown when the execution has a Claude Code session ID. Displays the full session ID in a code block for correlation with other logs.

Component Structure

skill-executions-content.tsx (list page)
  |-- SummaryCards (4 KPI cards + skill breakdown)
  |-- ExecutionsFilterBar (skill + status dropdowns)
  |-- Paginated executions table with links to detail and plan

skill-execution-detail.tsx (detail page)
  |-- ExecutionMetricsCards (4 cards)
  |-- ExecutionErrorCard (conditional)
  |-- MetadataViewer (JSON pre block)
  |-- Linked Research Plan card (conditional)
  |-- ExecutionAuditLog (session-filtered timeline)
  |-- Session ID card (conditional)

Related Pages

• Admin Dashboard Overview -- all admin views and shared patterns
• Research Plans View -- linked from the plan card on execution details
• Authorization -- Router Enforcement -- CASL checks on ResearchPlan subject