doctor

name	doctor
description	Run OmG diagnostics for extension readiness, workflow state integrity, and multi-agent safety.

Run OmG doctor diagnostics.

Input: $ARGUMENTS

Supported scopes:

• default (or status): quick extension health check
• team: multi-agent/team-pipeline readiness diagnostics
• workspace: lane/worktree hygiene and drift diagnostics
• hooks: hook-policy safety and trigger-health diagnostics

Diagnostic protocol:

1. Detect requested scope and run matching checks.
2. Verify extension surface integrity:
   • key directories (agents, commands/omg, skills, context)
   • key manifest/context files (gemini-extension.json, GEMINI.md, context/omg-core.md)
3. Verify command readiness for core flows (team, intent, workspace, taskboard, loop, rules, memory, hooks, notify, reasoning, approval).
4. Verify retained deep-work skills (plan, omg-plan, execute, prd, ralplan, research, deep-dive, context-optimize) and their metadata integrity.
5. Check state artifacts and drift risks under .omg/state/ when available, including lane health, trust notes, and hook lifecycle policy.
6. Detect common failure patterns:
   • extension not loaded in current session
   • Gemini CLI runtime too old for current OmG assumptions (recommend v0.38.0+)
   • missing or stale root GEMINI.md import chain
   • mixed or stale OmG extension roots (for example: copied assets, duplicate local clones, or stale installed extension contents diverging from the active repository)
   • command discovery mismatch
   • manual hook or skill wiring shadowing extension-managed OmG assets
   • malformed or duplicate skill metadata (skills/*/SKILL.md frontmatter)
   • native /memory inbox or skill-patching proposals applied without review or metadata validation
   • Plan Mode skill activation requested without explicit user confirmation
   • stale subagent guidance that assumes legacy wrapped subagent tools instead of the current unified invocation handoff
   • agent MCP server configs using an auth block without a documented credential boundary
   • stale/missing state files
   • taskboard entries with missing/null priority causing nondeterministic next selection
   • outdated depth flags (--essential, --standard, --deep) used instead of keywords (low, medium, high)
   • taskboard/workspace drift causing repeated replanning
   • missing lane baseline branch/HEAD anchors for active multi-lane work
   • baseline drift between workspace/taskboard anchor and the active branch/HEAD during resume or execution handoff
   • execution requested without confirmed team-plan + team-prd readiness artifacts
   • dirty shared worktree or untrusted lane being reused for review/autonomous flow
   • missing native worktree/sandbox capabilities expected by recent stable Gemini CLI updates
   • missing lane/subagent handoff context causing approval or verification drift
   • no fallback route when assigned execution/review agent is unavailable
   • workflow-order drift (execution requested before plan/prd acceptance on non-trivial tasks)
   • stale skill-active markers left behind after stop/cancel
   • missing or stale .omg/state/cancel-signal.json during resume handoff
   • hook safety policy gaps for autonomous loops
7. Produce concrete remediation actions in priority order.
8. Prefer extension-managed recovery paths first:
   • gemini extensions list
   • reinstall/update via gemini extensions ... when the active extension root looks stale
   • keep one authoritative OmG hook registration path per event before editing shipped assets
9. If filesystem tools are available, write/update .omg/state/doctor.md.

Response:

• Keep it concise and operator-facing.
• Include Doctor Result, Findings, and Recommended Next Command.
• Use the findings table only when there are multiple discrete issues.