The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands with administrative privileges and modifies system directories.
Evidence: Use of sudo apt-get install -y default-jre to install dependencies.
Evidence: Creating symlinks in /usr/local/bin/ to expose the binary globally.
[EXTERNAL_DOWNLOADS]: The skill downloads software from a third-party GitHub repository.
Evidence: Fetches the latest release version metadata from https://api.github.com/repos/AsamK/signal-cli/releases/latest.
Evidence: Downloads the binary archive from https://github.com/AsamK/signal-cli/releases/download/.
[REMOTE_CODE_EXECUTION]: The skill installs and runs a downloaded binary as a persistent background process.
Evidence: Extraction and execution of signal-cli tarball content.
Evidence: Running signal-cli in daemon mode as a child process.
[PROMPT_INJECTION]: The skill introduces a surface for indirect prompt injection by processing external message data.
Ingestion points: src/channels/signal.ts parses incoming JSON-RPC notifications from the Signal network.
Boundary markers: None; instructions do not specify any delimiters or safety warnings for the agent when processing message content.
Capability inventory: The skill has access to sensitive tools including run_code, write_file, and read_file.
Sanitization: No sanitization of message content or validation of incoming commands is documented.

add-signal