qruns

Audits scheduled runs and subagents from local Claude Code, Codex, and Cursor transcripts.

Updated Aug 14, 2026
One-click install
npx skills add https://github.com/QSchlegel/session-viz --skill qruns-qschlegel
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: qruns
Source: https://github.com/QSchlegel/session-viz/tree/main/plugins/session-viz/skills/qruns
Command: npx skills add https://github.com/QSchlegel/session-viz --skill qruns-qschlegel

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Autonomous agents and scheduled jobs often report success while silently shipping nothing, and vendor dashboards never surface this. This Skill reads every local transcript and shows which recurring tasks are blocked, silent, or quietly doing nothing, along with their token cost. ## Core Features & Use Cases - Delivery ledger: Lists every scheduled run, subagent, and workflow agent on the machine with terminal state, denied permissions, and write outcomes. - Failure classification: Marks tasks as BLOCKED (write refused by permissions), SILENT (intended write never landed), or quiet (correctly decided there was nothing to do). - Cost accounting: Breaks down token spend including cache-read per subagent family, surfacing invisible child runs. - Use Case: A nightly cron job has run 20 times with zero delivered output. Run the audit to see it is BLOCKED on a permission prompt, then fix it with a settings line via /qdoctor. ## Quick Start Ask the assistant to run /qruns and show which of my scheduled tasks and subagents are failing to deliver anything.

Frequently Asked Questions about qruns

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I check if my scheduled Claude Code tasks are actually delivering output?▼

Run /qruns to audit every scheduled run and subagent on your machine. It classifies each recurring task as BLOCKED, SILENT, or quiet, and shows run counts, terminal states, and token cost so you can see which jobs ship nothing.

Why does my headless agent run die without any error?▼

Headless runs cannot answer permission prompts, so they terminate at the first write attempt. Check the denied column in the qruns output, then run /qdoctor on that repo to add the missing allow entry to your settings.

What is the difference between BLOCKED, SILENT, and quiet tasks?▼

BLOCKED means a write was refused by permissions and a settings line fixes it. SILENT means an intended write never landed with no refusal, requiring transcript inspection. Quiet means the run correctly decided there was nothing to do, which is not a problem.

Does qruns verify that written files actually exist on disk?▼

No. The wrote_ok flag is a tool-result observation only; no filesystem probe runs in this audit. A successful write result and actual artifact presence remain separate facts, so do not treat wrote_ok as confirmed delivery.

Which coding agents and transcript formats does the audit support?▼

It reads local transcripts from Claude Code, Codex, and Cursor. The Cursor root is a single multi-gigabyte SQLite database rather than a file tree, and a full scan of a few thousand runs takes tens of seconds.

Why does cost per delivered show undefined for some tasks?▼

When a task delivers nothing, the denominator is zero, so the report prints undefined with both raw numbers instead of infinity or N/A. This refusal to fabricate a ratio keeps zero-delivery tasks honest and visible.