The Agent Skills Directory

[COMMAND_EXECUTION]: The skill orchestrates the codex CLI tool to perform code generation, refactoring, and analysis. It utilizes powerful flags such as --full-auto for autonomous file modification and --sandbox danger-full-access for tasks requiring network or system-level permissions.
[EXTERNAL_DOWNLOADS]: The skill configures a Model Context Protocol (MCP) server by downloading the @upstash/context7-mcp package via npx. This dependency is used to fetch and query external documentation from OpenAI and other sources.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection, as it is designed to ingest and process external content which may contain malicious instructions.
Ingestion points: Untrusted data enters the agent context through file reading (e.g., src/auth/login.py), git diff outputs, and linting error reports interpolated into execution prompts.
Boundary markers: The instructions do not demonstrate the use of clear delimiters or 'ignore' instructions to prevent the agent from following commands embedded within the processed data.
Capability inventory: The agent has access to sensitive capabilities including filesystem writes and system-level execution via the codex binary.
Sanitization: There is no documented mechanism for sanitizing or validating external data before it is included in prompts for the model.

codex-cli