skills.rest
hAcKlyc
hAcKlycCOMMUNITY
Explore all skills by this author →
GitHub

myagents-memory-molt

Audits and rewrites long-term workspace memory through structured self-reflection cycles.

863|101|Updated Jan 23, 2026
Search Tags:#self-reflection#introspection#long-term memory#memory audit#agent workspace#belief revision#Knowledge management
One-click install
npx skills add https://github.com/hAcKlyc/MyAgents --skill myagents-memory-molt
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: myagents-memory-molt
Source: https://github.com/hAcKlyc/MyAgents/tree/main/bundled-skills/myagents-memory-molt
Command: npx skills add https://github.com/hAcKlyc/MyAgents --skill myagents-memory-molt

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Long-term agent memory accumulates stale beliefs, contradictions, and outdated facts over time. This Skill performs a periodic deep audit (every 14 days) that rewrites SOUL, USER, and MEMORY files based on evidence, rather than just tidying notes.

Core Features & Use Cases

  • Belief Audit: Challenges every long-term judgment in MEMORY files against current evidence, downgrading or deleting beliefs that no longer hold.
  • External Fact Verification: Searches the web to verify time-sensitive claims (model versions, product statuses, market facts) and marks unverifiable ones explicitly.
  • Identity Coherence Check: Compares each SOUL principle against actual recent behavior to decide whether to revise the principle or the behavior.
  • Use Case: An agent workspace has been running for months and its MEMORY file contains outdated assumptions about model capabilities and user preferences. Run this Skill to produce a molt document, update the meta-layer files, and commit the changes to git.

Quick Start

Run a memory molt on this workspace to audit my long-term beliefs and update the SOUL, USER, and MEMORY files.

Frequently Asked Questions about myagents-memory-molt

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I audit and update long-term agent memory files?▼

Run the prepare_molt.py script with --repo pointing to your workspace to collect recent logs, topics, and previous molt promises. Then use --init-progress to generate a checklist, complete the five cognitive moves, and directly edit the SOUL, USER, and MEMORY files.

What is the difference between memory gardener and memory molt?▼

The gardener runs every 72 hours and handles pruning and demotion of memory content. The molt runs every 14 days and performs belief audits and self-rewriting of SOUL, USER, and MEMORY files, requiring at least one belief to be rejected or upgraded.

How often should a memory molt run?▼

The molt layer operates on a 14-day cadence, sitting above the 24-hour session capture layer and the 72-hour gardener layer. Running it more frequently reduces the cross-time evidence available for pattern mining.

Does the molt process require external web search?▼

Yes, external verification is a hard requirement for time-sensitive claims such as model versions, pricing, and product statuses. Claims that cannot be verified must be explicitly marked as unverified rather than asserted.

What files is the molt allowed to modify?▼

It may modify SOUL, USER, and MEMORY rule files under .claude/rules, plus files under memory/molts and the gardener flags file. Topic files are generally not rewritten during a molt.

Related Skills

Explore More Personal & Entrepreneur Skills→
GACLoveGACLove
2

openclaw

Organize AI context and memory through structured markdown files.

Community
Advanced
ghostwrightghostwright
1.5k

overheard

Monitor user promises from conversations and check follow-through.

Official
Intermediate
anticorrelatoranticorrelator
11

renormalize

Prunes stale entries, merges redundancies, and rebalances categories in a knowledge store.

Community
Advanced
NominexHQNominexHQ
15

pmm-init

Initialize structured git-backed memory with guided configuration and scaffolding.

Official
Advanced