The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to execute arbitrary shell commands via a bash tool. It provides extensive instructions on using pseudo-terminals (PTY) and background processes to run interactive command-line interfaces.
[REMOTE_CODE_EXECUTION]: The skill facilitates the execution of code generated by external AI agents. It explicitly encourages the use of the --yolo flag with the Codex CLI, which the documentation describes as 'most dangerous' because it bypasses all approvals, allowing the agent to execute actions and modify the workspace without human oversight.
[EXTERNAL_DOWNLOADS]: The skill recommends the installation of an unverified third-party package (@mariozechner/pi-coding-agent) from a personal repository/registry rather than a well-known service or official organization.
[PROMPT_INJECTION]: The skill is highly vulnerable to indirect prompt injection (Category 8) through its automated Pull Request review workflows.
Ingestion points: Untrusted code enters the environment via git clone and gh pr checkout commands documented in SKILL.md.
Boundary markers: No boundary markers or 'ignore' instructions are used to separate untrusted code from the agent's internal logic.
Capability inventory: The skill utilizes a bash tool capable of arbitrary command execution, file system modification, network access, and background process management (described in SKILL.md).
Sanitization: No sanitization or validation of the external code is performed before it is processed by the coding agents.

coding-agent