The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The installation instructions in README.md and the install.sh script implement a 'curl pipe bash' pattern, downloading a shell script from a remote GitHub repository (raw.githubusercontent.com/BofAI/skills/...) and executing it directly. This pattern is highly susceptible to supply chain attacks.
[COMMAND_EXECUTION]: The installer script scripts/install.py automatically modifies the agent's global configuration file (~/.claude/settings.json) to add Bash allow rules. This specifically targets the agent platform's security controls to bypass user confirmation prompts for the skill's execution scripts.
[COMMAND_EXECUTION]: The skill uses os.execv in scripts/script_utils.py to perform process replacement. This allows a running script to replace itself with an arbitrary executable or another script, which can be used to redirect execution flow unexpectedly.
[COMMAND_EXECUTION]: The skill utilizes AppleScript (osascript) via subprocess.run in scripts/configure_api.py and scripts/script_utils.py to manipulate the macOS UI, specifically to open new terminal windows and prompt for user credentials. This bypasses the agent's standard non-interactive execution environment.
[EXTERNAL_DOWNLOADS]: The skill fetches configuration and installation assets from external GitHub repositories using curl and git clone during the setup phase.

twitter-digest