The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the @openai/codex package from the npm registry. This originates from a well-known service and trusted organization (OpenAI).
[COMMAND_EXECUTION]: The skill documents native features of the Codex CLI that enable shell command execution, specifically the !command shortcut within the TUI and the exec subcommand for non-interactive automation.
[COMMAND_EXECUTION]: The instructions describe high-privilege flags and sandbox settings, such as --dangerously-bypass-approvals-and-sandbox (alias --yolo) and danger-full-access, which permit the agent to perform unrestricted operations on the host system.
[INDIRECT_PROMPT_INJECTION]: The skill describes an AI agent architecture designed to ingest project source code and user-provided instructions, which is a standard attack surface for indirect prompt injection.
Ingestion points: The tool reads files within the project directory and processes natural language prompts from the user.
Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded commands within the ingested project data.
Capability inventory: The tool possesses capabilities for filesystem modification (workspace-write), network access, and arbitrary shell command execution.
Sanitization: No explicit sanitization or input validation logic is described in the skill instructions.

codex-cli