From smorch-ops
SOP-11 enforcement. Default to Claude Code for all architecture + planning + review work. Codex (GPT) only with Mamoun's explicit approval and logged audit. This skill prevents silent Codex usage that introduces tool-mix inconsistency.
How this skill is triggered — by the user, by Claude, or both
Slash command
/smorch-ops:codex-doctrineThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
**Purpose:** SMOrchestra standardizes on Claude Code. Codex is not banned — it's gated.
Purpose: SMOrchestra standardizes on Claude Code. Codex is not banned — it's gated.
Before invoking Codex:
docs/codex-audit.csv:
date,task,claude_estimate,codex_estimate,actual_cost,outcome
/smo-plan checks current task: if task description matches Codex-tempting patterns but no audit row exists → warns/smo-score Engineering hat Q5 (elegance): if Codex was used without audit, pause skipped → caps Engineering at 7/smo-retro aggregates codex-audit.csv monthly: consistent Codex use for non-mechanical work → SOP reviewThis skill is NOT invoked as a command. It's a policy enforced at scoring + planning time.
When a PR's commit messages or PR description mention "used Codex for X":
/smo-score checks docs/codex-audit.csv for a matching entryDefault to Claude Code. Question every Codex urge. Document the cases where Codex wins. Over time, we should see Codex usage DECREASE as our Claude Code workflows get tighter (skills do more, prompt engineering reduces token spend).
SOP-07 (Agentic Coding Orchestration) defines the Claude Code vs Codex routing logic. This skill is the runtime enforcement of that policy. See SOP-07 for the full decision tree.
docs/codex-audit.csv: audit logProvides CDSS development patterns for drug interaction checking, dose validation, clinical scoring (NEWS2, qSOFA), and alert classification integrated into EMR workflows.
npx claudepluginhub smorchestra-ai/smorch-dev --plugin smorch-ops