OMC (Yeachan-Heo/oh-my-claudecode) cherry-pick — adopt Tier 1 + 2, reject Tier 3 (plugin install)
Decision
Analyzed Yeachan-Heo/oh-my-claudecode (OMC) against our setup. Cherry-picked 5 high-value capabilities as adaptations (not plugin install). REJECTED direct OMC plugin installation due to catastrophic conflicts with our existing Superpowers v5.0.5 pipeline + 13-hook architecture + MEMORY.md / CLAUDE.md / .claude.json canonical sources. Adopted: (1) Commit trailers protocol (Constraint/Rejected/Directive/Confidence/Scope-risk/Not-tested) — feedback_commit_trailers.md + global CLAUDE.md § COMMIT DISCIPLINE. (2) ast-grep CLI + skill wrapper — tree-sitter structural search/replace. (3) 4 user-scope specialized agents: critic (opus, adversarial), architect (opus, design), tracer (sonnet, debugging), scientist (sonnet, research). (4) ai-slop-cleaner skill — Phase 3.5 Polish on AI-authored code. (5) Serena MCP (oraios/serena v1.1.1) — IDE-grade LSP semantic code intelligence across 40+ langs, registered as Permanent #11. Rejected: OMC plugin (conflicts), OMC team pipeline (regression vs Superpowers v5.0.5), OMC state management (Vault/Graphiti richer), tri-model CCG (premature complexity), OMC HUD (our visible-checkpoint protocol is equivalent), OMC benchmarks (substantial infra, qualitative evidence sufficient).
Rationale
Our existing Superpowers v5.0.5 pipeline + Full Protocol + Vault/Graphiti compounding loop + Pristine Sweep is architecturally tighter than OMC’s team pipeline. OMC has broader agent library (19 vs our prior 6) + LSP integration (which we lacked) + commit trailer discipline (which we lacked). Cherry-picking keeps our architectural strengths (13 Laws, Tier 1 Directives, Vault cascade, dual-config, Hookify enforcement) while adopting OMC’s coverage gaps. Direct plugin install would overwrite our CLAUDE.md shape, introduce a competing team pipeline, and add a ~1MB TypeScript bridge for tmux workers we don’t need (we have Agent tool for subagent dispatch).
Alternatives Rejected
- Install OMC plugin directly — REJECTED (catastrophic conflicts). 2. Skip OMC entirely — rejected: lose 5 high-value capabilities (commit trailers, ast-grep, specialized agents, ai-slop-cleaner, LSP). 3. Adopt OMC’s whole approach including team pipeline — rejected: regresses our 2-stage review mandate, abandons TDD gate, loses brainstorming-first discipline. 4. Partial cherry-pick at Tier 1 only (3 items) — rejected: Tier 2 (Serena + ai-slop-cleaner) passed cost/benefit for current workload.
Outcome
Pending