The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill clones a remote repository from https://github.com/affaan-m/everything-claude-code.git to the local /tmp directory. This source is not a recognized trusted vendor and its contents are not verified before installation.
[REMOTE_CODE_EXECUTION]: The skill facilitates the installation of remote content into ~/.claude/skills/ and ~/.claude/rules/. Since these directories are automatically monitored and loaded by the agent environment, this allows an external repository to inject instructions or code that will be executed in future agent sessions.
[COMMAND_EXECUTION]: Uses several shell commands (git clone, rm -rf, cp -r, mkdir -p) to manipulate the file system. It specifically targets the ~/.claude/ directory, which is the sensitive core configuration path for the agent environment.
[DATA_EXFILTRATION]: The skill performs automated scans (grep) across all installed skill and rule files, including those in the user's home directory. While no direct network exfiltration was detected, the broad access to agent configuration files combined with network capabilities (git) creates a significant exposure surface.

configure-ecc