doctor_skill

Overview

This skill runs OrchestKit health diagnostics to validate plugin, agent, skill, hook, memory, and build integrity. It produces human-readable and machine-readable reports to help you catch configuration, schema, or runtime issues before they impact a team environment. Use it as a quick gate for CI, local troubleshooting, or post-install checks.

How this skill works

The doctor inspects installed OrchestKit plugins and runs a suite of checks across ten categories: installed plugins, skills, agents, hooks, permission rules, schema compliance, coordination locks, context budget, graph memory, and Claude Code version. It scans plugin and source folders, validates frontmatter and JSON schemas, parses JSONL memory files, and reports token usage and coordination lock state. Results can be emitted as plain text, verbose logs, or JSON for CI.

When to use it

• After installing or updating OrchestKit to verify integrity
• When hooks or background tasks are not firing as expected
• Before deploying to a shared or production environment
• When debugging coordination or lock-related failures across worktrees
• As part of CI pipelines to enforce build and schema consistency

Best practices

• Run /ork:doctor --json in CI to fail builds on regressions
• Use the -v/--verbose flag locally to get detailed failure traces
• Address frontmatter and schema warnings immediately to avoid runtime errors
• Verify .claude/memory/ decisions.jsonl lines parse as JSON when memory issues appear
• Rebuild plugins (npm run build) if build-system checks show mismatch

Example use cases

• CI job: run /ork:doctor --json and exit nonzero on any failed check
• Developer: run /ork:doctor -v after adding new skills or agents
• Ops: validate coordination and stale locks before scaling instances
• Troubleshooting: inspect memory graph integrity when subagents misbehave
• Pre-release: confirm Claude Code version meets the minimum runtime requirement