A Codex Skill for lightweight acceptance review and bounded fix loops.

Pi-native workflow baton runner with per-step model switching and isolated step context.

Quota-free iterative code review with continuous loop - architecture, security, OWASP, correctness, readability, performance, code quality, review loop

Claude builds in-session (no claude -p); Codex reviews headless. Every claimed result is re-run and diffed (claimed==actual) so hallucinated numbers die at the gate. A self-healing 6-stage build/review loop that runs until a real problem stops it.

Reusable Codex and Claude agent skills for VRM avatar polish: Blender/MMD inspection, SpringBone hair and skirt physics, clipping QA, backups, and evidence-driven review loops.

Quota-free iterative code review with continuous loop, architecture/security/OWASP/correctness/readability/performance review, ranked critical→nitpick with fix + AI prompt — idempotent .forge-standard-review/ per-repo memory

Privacy-first local MCP hub for coordinating Claude, OpenAI, Gemini, Ollama, and optional cloud council seats.

Fallibilist Plan → Implement → Review → Fix → Loop skill for robust agent workflows

Spec-gated, dual-review-loop implementation workflow for Claude Code