SkillOpt for ClaudeClaw Agents
Implement Microsoft Research's SkillOpt concept for ClaudeClaw's war room agents: treat each agent's procedural memory file as trainable external state that self-improves across runs.
- ▸skills/skillopt.md — Claude Code skill, invoke with /skillopt <agent>
- ▸scripts/gather-signals.ts — Queries ClaudeClaw's DB for real agent run outcomes
- ▸scripts/diff-and-propose.ts — Produces bounded unified diffs; enforces the 3-hunk limit
- ▸docs/integration-guide.md — Step-by-step wiring for ClaudeClaw-native and generic setups
- ▸docs/ground-truth-sources.md — Signal source reference table for all 16 war room agents
- ▸docs/scope-and-limits.md — Honest scope documentation and known failure modes
- ▸examples/scout-before-after.md — Real SCOUT optimization walkthrough with actual production diff
## goes-well-with
From the same shelf
Active Directory Security Audit Kit for Solo IT Teams
A fillable AD audit checklist covering privileged access, stale accounts, GPO hygiene, MFA gaps, and event log review — everything an auditor expects, packaged for the sysadmin who has no security team.
Agent Memory Architecture Part 2: Episodic vs Semantic vs Procedural Memory
The three agent memory types — episodic, semantic, and procedural — explained through the real production code of ClaudeClaw, a 24/7 multi-agent system, so your agents stop starting from scratch every session.
AI Chatbot Account Recovery: Security Audit Checklist for Developers
A developer's guide to designing safe AI chatbot account recovery flows, using the Meta/Instagram breach as a live case study.
## not-ready-to-buy
Take the field notes instead
One practical write-up a week from the same workbench these kits come from — plus reader pricing when new kits ship.