learning-system
ReviewAudited by ClawScan on May 10, 2026.
Overview
This instruction-only learning skill is not clearly malicious, but it asks to build cross-project memory and change future agent workflows without clear storage, approval, or rollback boundaries.
Install only if you want a persistent learning layer. Before using it, decide where its knowledge base lives, what data it may record, whether it can share learning across projects, and whether workflow or prompt changes require your approval.
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.
Sensitive project context, user feedback, or incorrect learned patterns could be carried into later tasks or other projects.
The skill explicitly plans to retain and reuse learning, profiles, and insights across projects, but the provided artifacts do not bound what is stored, how long it is kept, or how project data is isolated.
Knowledge Accumulation ... Cross-project insights ... KNOWLEDGE BASE ... Agent profiles ... Improvement history
Use only with explicit per-project opt-in, redaction of sensitive details, clear storage and retention rules, and a way to inspect, correct, and delete learned records.
A bad or poisoned learning record could change future agent behavior or involve additional agents without the user noticing.
The skill is designed to apply learned conclusions to future prompts, workflows, thresholds, and agent selection automatically, without visible approval, rollback, or containment requirements.
System Evolution - Agent prompt optimization - Workflow improvement - Quality threshold adjustment ... Automatically include /security-auditor in workflow when trigger conditions are detected.
Treat learned improvements as suggestions by default, require user confirmation before changing prompts or workflows, keep a changelog, and provide an easy rollback path.