- Entity format: 3-line compact registry (### Name / key facts / status+links) Heavy entries promoted to thread files. Cross-domain pointers for shared entities. - Pattern satellites: core patterns.md cap reduced from 110→70 lines (5.5KB). Domain-specific patterns go in satellite files loaded only by owning skill. - Reflect: pattern routing rules, entity format enforcement (step 3b) - Housekeeping: entity registry enforcement (step 5b), glacier inactive entities - Evolve: scorecard metrics for pattern distribution + entity compression ratio - README: updated entity example to 3-line format Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
3.6 KiB
Use this skill for systems-level self-improvement. Trigger if the user says "evolve", "system audit", "audit yourself", "check your architecture", or similar structural introspection requests.
This is NOT /reflect. Reflect = "what did I learn from interactions?" Evolve = "are the rules and architecture working?" Evolve never touches memory content — it changes the rules that govern how content moves.
Domain
Systems architecture — process rules, skill design, tier effectiveness, pipeline health.
Memory Files
Read FIRST — this is your continuity:
memory/cog-meta/evolve-log.md— your run logmemory/cog-meta/evolve-observations.md— architectural issues spotted
Architecture reference:
CLAUDE.md— project instructions.claude/commands/housekeeping.md— housekeeping rules.claude/commands/reflect.md— reflect rules
Measure (don't edit content):
memory/hot-memory.mdmemory/cog-meta/patterns.md- Any domain satellite pattern files (e.g.
work/*/patterns.md)
Process
1. Architecture Review
Evaluate the structural design:
- Tier design — are the tiers (hot-memory → patterns → observations → glacier) well-defined?
- Condensation pipeline — is the flow working? Where does it leak or stall?
- File naming and organization — any files in wrong domains? Orphaned files?
- Skill boundaries — are housekeeping/reflect/evolve boundaries clean? Any drift?
2. Process Effectiveness Audit
Review the output of recent housekeeping and reflect runs:
Housekeeping rules check:
- Did pruning priority order work? Or did it trim wrong things?
- Are glacier thresholds (50 obs, 10 action items) right?
- Is the 50-line hot-memory cap appropriate?
- Is entity format enforcement catching violations?
Reflect rules check:
- Did condensation produce useful patterns, or noise?
- Did thread candidate detection work?
- Is reflect staying in its lane?
- Are patterns routing to the right file (core vs satellite)?
Scorecard metrics — measure and record in evolve-log:
- Core
patterns.md: line count / 70, byte size / 5.5KB (target: ≤1.0) - Satellite pattern files: list each with line count (soft cap: 30)
- Entity compression ratio:
(total entity lines across all files) / (total ### entries)(target: ≤3.0) - Hot-memory line counts vs caps
3. Rule Change Proposals
Based on findings, propose concrete rule changes. Don't fix content — fix the rules.
For each proposal:
- What problem does it solve?
- What evidence supports it?
- What's the risk?
Apply rule changes directly to the relevant skill files if clearly beneficial and low-risk. For changes that affect user-facing behavior, note them as proposals for user review.
4. Write Observations & Update Log
Observations — Append to memory/cog-meta/evolve-observations.md:
- Format:
- YYYY-MM-DD [tag]: observation - Tags: bloat, staleness, redundancy, gap, architecture, opportunity, rule-drift, process-health
Evolve Log — Append to memory/cog-meta/evolve-log.md:
- Run number, process effectiveness findings, rule changes applied or proposed, deferred items
- Update "Next Run Priorities" section at top
5. Debrief
Concise summary:
- Process health — did housekeeping/reflect follow their rules?
- Rule changes — applied or proposed, with rationale
- Architecture notes — structural observations
- Next evolve — top 3 things to check next time
Keep it actionable. Numbers over narrative.
Activation
Read evolve-log.md and evolve-observations.md FIRST for continuity. Then audit the system. You are the architect — you design the rules, you don't play by them.