The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands to interact with the Codex CLI, including codex exec, codex login, and codex --version. It also references executing local scripts via uv run and make. These commands are consistent with the skill's stated purpose of delegating tasks to an external CLI tool.
[EXTERNAL_DOWNLOADS]: The skill provides instructions to install the @openai/codex package globally using npm. This package is an official tool from OpenAI, a well-known and trusted service provider, and its inclusion for the purpose of the skill is appropriate.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied prompts as positional arguments for the codex exec command. This represents a potential surface for indirect prompt injection if the inputs are not properly sanitized or if the agent executes instructions embedded within the data provided to the Codex model.
Ingestion points: Prompts are passed as positional arguments to codex exec or as a PROMPT variable in a Makefile.
Boundary markers: No explicit boundary markers or delimiters for untrusted data are shown in the usage examples.
Capability inventory: The skill can execute shell commands via the CLI, write output to files in the delegations/codex/ directory, and read file context from the local filesystem.
Sanitization: There is no evidence of sanitization or escaping of the prompt input within the provided documentation.

codex-delegation