Working-tree (uncommitted diff) code review with cross-model adversarial pass — Gemini reviews, Codex challenges false positives, surviving findings get applied. Use BEFORE commit. For PR-level review, use the official `/code-review:code-review` plugin instead.
Review CORE level plans (A1-C2, PRO) using State Standard 2024 + textbook RAG + VESUM. Finds bad vocab, wrong grammar scope, Russianisms, factual errors.
Atomic rebuild for Core B (B1 M06+, B2, C1, C2, PRO). Narrative Engine v5.0 (Slim Skill + Rich Phase Prompts).
Adversarial final review for otaman-built modules. Runs in a NEW session (separate from builder/reviewer) to maintain independence. Catches what automated audit and Green Team missed.
Validate pipeline prompt templates. Finds unreplaced placeholders, contradictions between templates, stale instructions.
Apply fixes from plan review reports. Reads audit reports, shows proposed changes for approval, then applies them. Respects plan versioning rules.
Audit prompt sizes across build orchestration dirs. Surfaces sections that dominate the prompt (>40%), tracks per-phase budget vs the model's context window, and checks whether directives in the prompt were actually covered in the response.
Review SEMINAR track plans (HIST, BIO, ISTORIO, LIT, OES, RUTH) using Wikipedia + Literary RAG + VESUM. Finds factual errors, ghost sources, decolonization issues.