The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to download and install unversioned executable binaries from an external third-party repository on GitHub (github.com/steveyegge/gastown) using the go install ...@latest command.
[PRIVILEGE_ESCALATION]: Setup instructions include the use of sudo to install the GitHub CLI (gh) on Linux systems.
[PERSISTENCE]: The skill directs the agent to modify the user's shell configuration files (.bashrc or .zshrc) to ensure the Go binary directory remains in the system PATH across sessions.
[COMMAND_EXECUTION]: The core functionality relies on the agent autonomously executing shell commands for the gt and bd CLI tools, explicitly instructing the agent to bypass user interaction and confirmation for these operations.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a significant attack surface for indirect prompt injection: 1. Ingestion points: External data enters through gt rig add (cloning GitHub repositories), gt mail (inter-agent messaging), and gt peek (monitoring output from other agents). 2. Boundary markers: No explicit delimiters or instructions are used to prevent the agent from following instructions embedded in external content. 3. Capability inventory: Extensive capabilities including shell command execution, file system modification, and network operations. 4. Sanitization: Lacks mechanisms to validate or filter the content of processed repositories, issues, or agent messages.

gastown