The Agent Skills Directory

[SAFE]: The skill's instructions focus on improving the quality and clarity of dialogue. No attempts to bypass safety filters, escalate privileges, or exfiltrate data were found. The use-case (question generation) is low-risk as the skill serves as a cognitive guide rather than a functional tool with system-level access.
[NO_CODE]: There are no scripts, binaries, or configuration files that trigger code execution. The entire skill is comprised of markdown text and a test evaluation JSON file, which significantly minimizes the potential attack surface.
[EXTERNAL_DOWNLOADS]: The README file provides a link to the author's public GitHub repository (github.com/synapseradio/ai-skills) for installation. This is a transparent and standard distribution method for the author's own work and does not involve the AI agent performing hidden or unsafe downloads.
[SAFE]: While the skill is designed to process untrusted user data (such as project specifications) to generate questions, it lacks the necessary 'capabilities' (such as shell access, file writing, or network operations) to be exploited via indirect prompt injection.

ask-questions