acpx
Warn
Audited by Gen Agent Trust Hub on Aug 24, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill registry specifies commands that launch various CLI tools as adapters and includes an escape hatch via the
--agentflag that allows for the execution of arbitrary shell commands. - [EXTERNAL_DOWNLOADS]: The skill instructions and registry utilize
npxanduvxto fetch and execute several third-party packages, including@agentclientprotocol/codex-acp,@kilocode/cli, andfast-agent-mcp. - [REMOTE_CODE_EXECUTION]: Several components of the agent registry utilize
npx -yanduvx, which facilitate the automatic download and execution of remote code at runtime without manual approval. - [PROMPT_INJECTION]: The skill contains instructions marked as 'CRITICAL' that explicitly override the agent's default behavior and internal examples to enforce a specific protocol using isolated SubAgents and independent result reflection.
Audit Metadata