megaplan-observe

Diagnose stalled or drifting megaplan runs via introspection and event timelines.

97|8|Updated Mar 20, 2026

Search Tags:#debugging #introspection #workflow state #trace analysis #LLM operations #megaplan #run recovery

Ready to run

npx skills add https://github.com/peteromallet/arnold --skill megaplan-observe

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: megaplan-observe
Source: https://github.com/peteromallet/arnold/tree/main/megaplan/data/_codex_skills/megaplan-observe
Command: npx skills add https://github.com/peteromallet/arnold --skill megaplan-observe

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill eliminates the guesswork when a megaplan run appears to be progressing slowly, getting stuck, or drifting from its intended plan by showing an actionable, structured view of in-flight state and events.

Core Features & Use Cases

In-flight introspection: Produces a single structured JSON payload via megaplan introspect --plan X for fast triage.
Live event tracing: Uses megaplan trace --follow to watch what the system is doing in real time.
Doctor-style diagnostics: Uses megaplan doctor --plan X to validate flags/states and provide remediation hints.
Failure-mode guided recovery: Interprets liveness, detect blocked states with recoverable_via, and flags rubric/binary profile drift via rubric_doc.drift.

Quick Start

Ask the Skill to inspect an active plan by running: megaplan introspect --plan my-sprint.

Frequently Asked Questions about megaplan-observe

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

FAQPage Schema

How do I debug a stalled or drifting megaplan run?▼

Debug a stalled megaplan run by diagnosing in-flight state and event timelines to determine liveness, detect recoverable transitions, and flag rubric or profile mismatches. This eliminates guesswork when runs get stuck or drift from their intended plan.

How does live event tracing work for an active workflow state?▼

Live event tracing uses `megaplan trace --follow` to watch system execution in real time during an active workflow state. It produces a structured view of events to show exactly what the system is doing and why.

What is the best way to introspect an active plan for fast triage?▼

The best way to introspect an active plan for fast triage is producing a single structured JSON payload via `megaplan introspect --plan X`. This yields an actionable, structured view of in-flight state and events for immediate diagnosis.

Can I detect rubric or binary profile drift during LLM operations?▼

Yes, you can detect rubric or binary profile drift during LLM operations by interpreting `rubric_doc.drift` via `megaplan doctor --plan X`. This validates flags and states to flag mismatches and provide remediation hints.

Why does my megaplan execution get blocked and how do I recover it?▼

Megaplan execution gets blocked when liveness fails or transitions stall, but you can recover it by interpreting `block_details.recoverable_via` using `megaplan introspect`. This identifies failure modes and guides recovery actions.