The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install a command-line tool from a third-party Homebrew tap and a GitHub repository that are not associated with a recognized trusted organization.\n
Evidence: brew install kylebrowning/tap/ble and mint install kylebrowning/BLESwiftCLI (SKILL.md).\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external Bluetooth Low Energy (BLE) devices, which could be used to influence the agent's behavior if the data contains malicious instructions.\n
Ingestion points: Data is received through commands such as ble scan (device names, advertisements) and ble read (characteristic values) (SKILL.md).\n
Boundary markers: The skill does not provide specific delimiters or instructions for the agent to ignore potentially malicious content within the Bluetooth data.\n
Capability inventory: The tool can perform write operations to devices and open L2CAP channels, and its output can be piped to other system commands like jq or shell scripts (SKILL.md).\n
Sanitization: There is no evidence of filtering or sanitizing the data received from external devices before it is processed by the agent.\n- [COMMAND_EXECUTION]: The skill relies on the execution of shell commands and the installation of a binary tool on the local system.\n
Evidence: Use of ble, brew, mint, and swift build commands in the installation and usage sections (SKILL.md).

bleswiftcli-macos-bluetooth