Security audit
Security checks for vulnerabilities and agentic risk
This skill is a knowledge-routing helper that can write notes or playbooks when asked, with no evidence of hidden execution, credential access, networking, or deception.
Before using it to execute writes, check the proposed files and draft content, especially when the task involved private details, credentials, or temporary debugging output. It is reasonable to install for knowledge-management workflows.
The skill’s trigger scope is very broad: it activates on essentially any post-task knowledge routing decision, which increases the chance of unintended invocation on sensitive or low-signal work. In practice, that can lead to over-collection, accidental persistence of transient or confidential task details, and unnecessary modification of repository knowledge artifacts.
The instruction to 'write the files directly' enables repository modification without any explicit user-facing warning, confirmation, or safety gate. That is dangerous because this skill manages long-lived knowledge stores, so a mistaken or manipulated invocation can create, overwrite, or persist incorrect, sensitive, or adversarial content into trusted documentation paths.
No suspicious patterns detected.
