๐ฉบ memory-health-check
AdvisoryAudited by Static analysis on Apr 30, 2026.
Overview
No suspicious patterns detected.
Findings (0)
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.
If this capability is actually granted, the agent could have authority to initiate paid actions or subscriptions unrelated to a memory health scan.
The purchase capability is not needed for local memory integrity, freshness, bloat, orphan, duplicate, or coverage diagnostics, and no explicit purchase approval flow is documented.
Capability signals - crypto - can-make-purchases
Do not grant purchase/billing authority to this skill unless a separate, explicit, user-confirmed purchase workflow is documented and autonomous use is disabled for that action.
A mistaken orphan classification or overly broad cleanup could delete useful local memory files.
The repair tooling is deletion-capable. It is purpose-aligned and has a dry-run mode, but memory cleanup can still remove entries the user may want to keep.
# Dry run (preview what would be deleted) python3 scripts/auto_repair.py --dry-run # Remove orphans (inbound-unlinked files) python3 scripts/auto_repair.py --remove-orphans
Run dry-run first, review the proposed deletions, and back up important memory data before using auto-repair.
Reports may expose metadata about private memories, file paths, database tables, duplicate relationships, or stale entries, even though no network exfiltration is shown.
The skill intentionally reads persistent OpenClaw memory stores, including other agents' memory paths, to produce diagnostics.
DISCOVER MEMORY PATHS
โข ~/.openclaw/workspace/memory/ (A-layer)
โข ~/.openclaw/memory/*.sqlite (B-layer)
โข ~/.openclaw/workspace/*/memory/ (other agents)Only run it where scanning these memory locations is acceptable, and review generated reports before sharing them.
A user or agent following the recommendation may run a separate skill with its own risks and permissions.
The skill may recommend commands from another skill that is not declared as an installed dependency in the supplied metadata. The shown code records recommendations rather than automatically executing them.
action="CRITICAL: Memory bloat is severe. Run dreaming-optimizer to consolidate entries.", cli_command="~/.openclaw/workspace/skills/dreaming-optimizer/bin/optimize.sh"
Review and approve the referenced dreaming-optimizer skill separately before running its commands.