agent-teams-playbook
Warn
Audited by Gen Agent Trust Hub on Mar 13, 2026
Risk Level: MEDIUMREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The installation script (
scripts/install.sh) fetches theSKILL.mdandREADME.mdfiles from a remote GitHub repository (KimYx0207/agent-teams-playbook) that is not listed as a trusted vendor. This repository is controlled by an individual rather than a known organization. - [REMOTE_CODE_EXECUTION]: The skill logic in
SKILL.md(Stage 1, Step 2) directs the agent to search for and install external skills using the commandnpx skills add <owner/repo@skill-name> -g -y. This pattern allows for the dynamic download and execution of arbitrary code from third-party sources encountered during the skill discovery process. - [COMMAND_EXECUTION]: The
scripts/install.shscript performs several local system operations, including creating directories (mkdir -p), removing existing directories (rm -rf), and setting file permissions (chmod +x). While common for installation scripts, these actions should be reviewed if coming from an untrusted source. - [PROMPT_INJECTION]: The skill uses strong terminology such as "铁律" (Iron Law) and "硬性标准" (Hard Standards) to enforce its own internal logic and workflow. While used here for orchestration consistency, such patterns can sometimes be used to prioritize skill-specific instructions over general agent safety guidelines.
Audit Metadata