# Self Optimization Loop

> Full NexusOS Self-Optimization Loop (SOL) procedure for continuous prompt improvement: discover, audit, build, approve, apply, track, and persist prompt optimizations. Use when Pafi says /sol, asks for SOL governance, or requests the complete continuous prompt-improvement workflow.

- Skill: `cryptopafi/self-optimization-loop` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add cryptopafi/self-optimization-loop`
- Raw SKILL.md: https://api.skillmd.com/api/skills/cryptopafi/self-optimization-loop/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: cryptopafi (https://skillmd.com/u/cryptopafi)
- Updated: 2026-09-21
- Page: https://skillmd.com/skills/cryptopafi/self-optimization-loop

---


# Self-Optimization Loop (SOL)

Use this skill when Pafi asks for the complete SOL procedure, SOL governance, SOL prompt optimization workflow, or implementation details for the NexusOS self-optimization loop.

This is distinct from `/sol-cycle`:
- `self-optimization-loop` is the complete SOL procedure and governance workflow.
- `sol-cycle` is the runtime skill for executing one queue item through audit, validation, approval, apply, and Cortex save.

## Canonical Procedure

Read and follow the complete imported procedure at:

`references/SELF-OPTIMIZATION-LOOP.md`

The reference file is an exact copy of the canonical NexusOS procedure from:

`/Users/pafi/.nexus/procedures/SELF-OPTIMIZATION-LOOP.md`

## Execution Rules

1. Before modifying prompts or SOL infrastructure, read the canonical procedure reference.
2. Use the procedure's phase model: Discover -> Audit -> Build -> Approve -> Apply -> Track.
3. Preserve schema, safety rails, approval gates, Cortex persistence, and manifest/queue behavior exactly unless Pafi explicitly asks for a procedure update.
4. For one-item runtime execution, delegate to the separate `/sol-cycle` workflow after confirming the queue/audit state.
5. If the canonical procedure and runtime skill disagree, treat the canonical procedure as policy and the runtime skill as implementation detail.

## Codex Model Routing

- Use `gpt-5.6-sol` with `max` reasoning for SOL audits, governance decisions, and final verification.
- Use `gpt-5.6-sol` with `high` reasoning for bounded prompt-building passes where the audit already supplies the required changes.
- Treat historical Opus/Sonnet actor names in the imported procedure as role labels when executing in Codex; preserve their authority separation even when both roles use GPT-5.6 Sol.