skill
Warn
Audited by Gen Agent Trust Hub on May 12, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The
/skill setupand/skill scansubcommands utilize shell scripts involvingfind,sh -c,grep, andsedto perform inventory and metadata extraction from markdown files in the~/.codex/skills/and.codex/skills/directories. - [EXTERNAL_DOWNLOADS]: The
/skill importfeature allows the agent to download skill content from arbitrary user-provided URLs. While GitHub Gist is mentioned as a common use case, the functionality is not restricted to trusted sources. - [REMOTE_CODE_EXECUTION]: There is a potential risk of code execution through the processing of imported skills. If a maliciously crafted skill is imported with shell metacharacters in its metadata (like the 'name' or 'description' fields), the subsequent shell-based scanning logic used in the inventory process could potentially execute unintended commands.
Audit Metadata