build-skill
Warn
Audited by Gen Agent Trust Hub on Mar 3, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The main
SKILL.mdfile instructs users to useuvxto download and run a tool fromhttps://github.com/agentskills/agentskills.git. This repository is not included in the trusted vendors list. - [REMOTE_CODE_EXECUTION]: The recommendation to execute remote code using
uvxwith a direct link to a Git repository (github.com/agentskills/agentskills.git) allows for the execution of unverified logic within the agent's environment. - [COMMAND_EXECUTION]: The skill bundles three shell scripts (
scripts/init_skill.sh,scripts/validate_skill.sh, andscripts/package_skill.sh) that perform local file system operations, create directory structures, and use system utilities likezipandfind. These scripts are used for skill management and scaffolding. - [DATA_EXPOSURE]: The
scripts/package_skill.shfile includes a proactive security measure by defining a list of sensitive file patterns (such as.env,.pem,id_rsa, andcredentials) to be excluded automatically when packaging a skill into a zip file, reducing the risk of accidental credential exposure.
Audit Metadata