Structured Memory

Manage and update a layered memory system using daily files as source, indexing by domain/module/entity, extracting critical facts, and maintaining recall ef...

This skill appears to do exactly what it says: parse your daily notes and build local indexes and critical-facts. It does not contact external servers or require credentials. Before enabling it: 1) Inspect your existing memory/YYYY-MM-DD.md files for any sensitive identifiers you would not want extracted into workspace files. 2) On first run, consider using --no-backfill or run init_structure.py manually so you control which days are processed. 3) If you are worried about sensitive data, search the skill bundle for where 'account' / 'host' / 'credential' facts are written (critical-facts/credentials.md, hosts.md, etc.) and adjust the code or your notes accordingly. 4) Optionally run the tests and a single-day rebuild on a copy of your workspace to see what would be written. If you want a more conservative installation, disable automatic backfill and review/clean daily notes before running the write-enabled extraction.

Type: OpenClaw Skill Name: structured-memory Version: 1.0.1 The structured-memory bundle implements an automated system for indexing and extracting technical metadata from an agent's daily logs. The script `scripts/extract_critical_facts.py` aggressively scrapes sensitive information including IP addresses, file paths, account names, and UUIDs using regex. While this is aligned with the stated purpose of technical memory maintenance, the `SKILL.md` file contains prompt-injection-style instructions that direct the AI agent to automatically execute these scripts as a 'default operating rule' whenever meaningful updates occur. The use of `subprocess.run` across multiple scripts to orchestrate these tasks, combined with the autonomous execution instructions and the collection of sensitive technical identifiers, presents a high-risk profile despite the lack of clear evidence of data exfiltration or malicious intent.