Security audit
Security checks for vulnerabilities and agentic risk
This skill is a simple self-improvement logger/analyzer with disclosed local logging, but users should understand it may analyze conversations and keep improvement notes in workspace files.
Before installing, decide whether you are comfortable with conversation-quality analysis and local retention of improvement notes. Use it only in workspaces where improvement_log.md can safely contain derived insights, and disable or avoid automatic analysis for sensitive conversations until the skill provides clearer consent, retention, and purge controls.
The skill description presents automatic conversation analysis and learning logs without warning users that conversation-derived data may be persistently stored. This is dangerous because users may disclose sensitive or regulated information under the assumption that the interaction is ephemeral, while the skill silently retains summaries or insights to disk.
The skill advertises behavior that reads and writes local files such as ./improvement_log.md and ./SOUL.md, but it does not declare any permissions or allowed-tools scope. This creates an authorization and transparency gap: a host may enable the skill without clearly constraining or surfacing its filesystem access, increasing the chance of unintended data exposure or tampering.
The statement that the skill 'automatically analyzes conversations after each session' is an overly broad auto-activation trigger with no scope limits, exclusions, or consent boundary. In practice, this can cause the skill to run on sensitive conversations, process data unrelated to improvement tasks, and persist outputs without the user intentionally invoking it.
The description claims a broadly capable 'self-improving agent system for continuous optimization' without defining boundaries, triggers, or prohibited behaviors. In agent ecosystems, vague self-improvement and optimization claims can enable unsafe invocation, scope creep, or downstream components granting excessive autonomy because the skill appears generally applicable.
The heuristic checks only Chinese and English keywords when interpreting user feedback, which embeds a language preference into the skill's behavior without offering user choice or documenting the limitation. This can disadvantage users providing feedback in other languages and constitutes a locale/language policy concern.
No suspicious patterns detected.
