auto-memory-curation_skill

This skill silently analyzes every message to extract important details and stores them in memory files for passive context building.
  • Python

2.5k

GitHub Stars

2

Bundled Files

2 months ago

Catalog Refreshed

3 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill openclaw/skills --skill auto-memory-curation

  • _meta.json292 B
  • SKILL.md3.5 KB

Overview

This skill automatically analyzes every incoming message and stores only meaningful information into appropriate memory files. It runs silently on each message to build context over time without manual prompts. The aim is passive, accurate memory curation that filters noise and preserves actionable facts, decisions, and preferences.

How this skill works

The skill runs on every message and first filters out noise such as greetings, thanks, single-word replies, and bot commands. For non-noise messages it classifies content into categories (facts, decisions, preferences, ideas, learnings, projects, goals, errors, commitments). It extracts a concise one-sentence memory with a timestamp and appends it to the right file or folder, ensuring no overwrites and checking for duplicates.

When to use it

  • When you want passive memory building without explicit "remember this" commands
  • When you need consistent capture of decisions and commitments across many conversations
  • To accumulate user preferences and learning insights over time
  • When tracking project updates, blockers, and milestones across messages
  • When you want a lightweight audit trail of corrections and anti-patterns

Best practices

  • Be selective: prefer storing high-value items (decisions, commitments, unique facts) over chatter
  • Write one concise sentence per memory with minimal paraphrasing to preserve accuracy
  • Include contextual metadata (date, short context line) so items are useful later
  • Avoid storing secrets: never capture passwords, API keys, or other sensitive data
  • Periodic review: audit recent memories to tune filters and prevent drift

Example use cases

  • Capture a user preference like communication style into USER.md for future responses
  • Record a project decision and append it to the dated project memory file
  • Save an insight or learning to memory/topics/lessons.md for later retrieval
  • Log a commitment or task into tasks.md so it can feed task tracking workflows
  • Archive recurring errors or anti-patterns into memory/topics/anti-patterns.md to avoid repetition

FAQ

Yes. Say commands such as "Don't store that" or "Clear recent memories" to prevent or remove stored items.

Does the skill ever store secrets or passwords?

No. The filter explicitly skips passwords, API keys, and other sensitive credentials.

How does the skill avoid duplicates?

Before appending, it checks existing memory entries for the same concise fact or decision and skips duplicates to keep records clean.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational