The Agent Skills Directory

[REMOTE_CODE_EXECUTION] (CRITICAL): The skill contains instructions for the agent to download a script and execute it immediately without verification.
Evidence: In SKILL.md, the command curl -fsSL "https://raw.githubusercontent.com/Dicklesworthstone/beads_rust/main/install.sh?$(date +%s)" | bash is provided as the recommended installation method.
Risk: This pattern allows an untrusted third party (the repository owner) to execute arbitrary code on the host system. The use of $(date +%s) in the URL is a cache-busting technique that ensures the most recent (potentially malicious) version of the script is always fetched.
[EXTERNAL_DOWNLOADS] (HIGH): The skill encourages downloading software from a non-whitelisted, untrusted source.
Evidence: References to github.com/Dicklesworthstone/beads_rust in both SKILL.md and README.md for tool installation.
Risk: Untrusted sources do not have the security guarantees of the platforms listed in the Trusted External Sources rule, increasing the risk of supply chain attacks.
[COMMAND_EXECUTION] (MEDIUM): The skill defines broad command execution permissions for the br tool.
Evidence: .claude/settings.local.json grants permission for Bash(br:*).
Risk: While necessary for the skill's functionality, these permissions facilitate the impact of the CRITICAL RCE finding by allowing the installed tool to perform wide-ranging actions across the filesystem.

beads-rust