Agent Memory Hierarchy
PassAudited by ClawScan on May 15, 2026.
Overview
This instruction-only skill is coherent for organizing agent memory, but users should be aware it encourages persistent storage of personal facts and local memory-file edits.
This skill appears safe to install as an instruction-only memory organizer. Before using it, decide what information should be allowed in persistent memory, avoid storing actual secrets or API keys, and review backups and migrated files so future agents do not rely on incorrect or overly sensitive context.
Findings (2)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Personal facts, account references, local paths, and behavioral context may become part of the agent's long-term memory and affect future responses.
The skill directs the agent to persist structured personal, account, system, and project facts across sessions. This is central to the memory-management purpose, but it means sensitive or incorrect data could be retained and reused later.
`facts.yaml` — All structured factual data: ... Identity facts (name, contact, IDs) ... File paths and system config ... Services and accounts ... After each session: update `facts.yaml` if any facts changed
Store only facts you actually want retained, avoid putting secret values in memory, review the generated files periodically, and remove stale or sensitive entries.
A migration mistake could remove, misclassify, or over-prioritize important memory entries.
The skill instructs the agent to read and modify local memory files. These edits are expected for the skill's purpose, but they can change what the agent remembers and how it behaves later.
Step 1: Audit existing MEMORY.md ... Step 4: Rewrite MEMORY.md ... When updating facts: Edit `facts.yaml` directly.
Back up existing memory files before migration and review MEMORY.md, facts.yaml, and lookup-index.md after changes.