syn-insights

name	syn-insights
description	Query Syntropic137 observability data; sessions, artifacts, token costs, tool timelines, and system-wide insights
argument-hint	<overview|sessions|artifacts|costs|tools|tokens> [args]
model	sonnet

/syn-insights: Observability

Use this skill when you need to answer two questions: "Why was this execution expensive?" and "Why did this session fail?" These are the questions observability data exists to answer.

When to Use This

Use /syn-insights when you want to: browse sessions and their costs, inspect what artifacts an execution produced, trace which tools a session called and in what order, or check platform-wide cost summaries.

For the conceptual model behind sessions, cost projections, and the two-lane architecture, the observability skill has the full explanation. For controlling a running execution, use /syn-control.

Sessions

A session = one Claude CLI invocation in one workspace. A 3-phase workflow creates 3 sessions.

syn sessions list                              # all sessions
syn sessions list --workflow <workflow-id>     # filter by workflow
syn sessions list --status running             # filter by status
syn sessions list <session-id>                 # session detail with operations log

API: GET http://localhost:8137/api/v1/sessions (supports ?workflow_id=<id>&status=<status>)

Artifacts

syn artifacts list
syn artifacts list --workflow <workflow-id>
syn artifacts show <artifact-id>
syn artifacts content <artifact-id> --raw

Costs

syn costs summary                              # platform-wide summary
syn costs session <session-id>         # cost breakdown for one session
syn costs execution <execution-id>     # cost aggregated across all phases

The breakdown includes cost_by_model and cost_by_tool; these tell you whether opus usage or expensive tool calls (like Bash) drove the cost.

Tool Call Timeline

The most useful debug view, showing every tool call in order with timing and success/failure:

syn observe tools <session-id>
syn observe tools <session-id> --limit 200

Look for TOOL_BLOCKED entries (the agent was denied a tool) and long duration_ms values (slow file reads, hanging bash commands).

Token Metrics

Per-session token breakdown including cache hit/miss:

syn observe tokens <session-id>

Low cache_read_tokens relative to input volume means the agent is re-reading large files every turn, which is expensive and slow.

Common Scenarios

"This execution cost $12. What happened?"

1. syn costs execution <id> to find which phase was most expensive
2. syn costs session <session-id> for the expensive phase: check cost_by_model and cost_by_tool
3. syn observe tokens <session-id> to check cache hit ratio

"The workflow produced no artifacts. What did the agent do?"

1. syn sessions list --workflow <id> to find the session IDs
2. syn observe tools <session-id> to trace every tool call; look for errors or blocked tools

"I want to see all sessions from the last run." syn sessions list --workflow <workflow-id> lists all sessions in order, most recent first.

Platform Overview

Cross-org health + cost summary: syn insights overview

This hits GET /api/v1/organizations/overview and shows total executions, costs, and system health status.

Errors

On API errors, run /syn-health. For the full observability model (two-lane architecture, event pipeline, cost model), see the observability skill.