The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides instructions to install the Google Apps Manager (GAM) tool using a dangerous execution pattern that downloads and runs remote code.
Evidence: bash <(curl -s -S -L https://gam-shortn.appspot.com/gam-install) in SKILL.md.
This pattern downloads a script from a remote URL and pipes it directly into a bash process substitution for immediate execution. This execution model bypasses local inspection and could allow for arbitrary code execution if the source or the transmission is compromised.
[PERSISTENCE]: The skill instructs the agent/user to establish a persistence mechanism using the system cron daemon.
Evidence: echo "0 */4 * * * curl -s -X DELETE ... | crontab -" in SKILL.md.
While the intended purpose is for periodic cleanup of temporary group memberships, the use of crontab creates a persistence layer where commands are executed automatically in the background across sessions.
[COMMAND_EXECUTION]: The skill relies extensively on complex shell command pipelines, including the interpolation of tokens and identifiers into subshells.
The workflows involve chaining multiple CLI tools (e.g., curl, jq, az, gam, gh, aws) and using command substitution (e.g., USER_OID=$(az ad user show...)). Such patterns carry a risk of command injection if variables originating from external inputs or API responses are not properly sanitized before being passed to the shell.

identity-access-management