The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's primary function is the execution of arbitrary shell commands via the exec_command and background_process tools to manage sub-agent lifecycles.
[REMOTE_CODE_EXECUTION]: The skill provides instructions to download and install third-party software from an unverified source (@mariozechner/pi-coding-agent). It also encourages dynamic dependency installation using pnpm install within project environments that may contain untrusted code.
[EXTERNAL_DOWNLOADS]: The skill fetches official CLI tools from Anthropic and OpenAI. It also facilitates cloning arbitrary external repositories from GitHub for analysis, refactoring, and review tasks.
[PROMPT_INJECTION]: The skill explicitly instructs the agent to use CLI flags that bypass standard security confirmations and sandboxing, such as the --yolo flag for Codex and --permission-mode bypassPermissions for Claude Code.
[PROMPT_INJECTION]: A significant surface for indirect prompt injection exists because the skill is designed to ingest and process untrusted data from external repositories and pull requests. A malicious file within a project could contain instructions that a sub-agent, operating with high autonomy and tool access, might execute on the host system. The instructions lack specific boundary markers or sanitization steps to mitigate this risk.

sub-agent