The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is entirely built around the execution of the codex command-line tool via Bash. It provides specific templates for invoking codex exec with varying reasoning efforts and sandbox permissions, including workspace-write capabilities.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the transfer of task data and instructions to a sub-agent. This process lacks explicit sanitization or boundary markers between user-controlled task data and the resulting command-line prompts.
Ingestion points: Task objectives, scope definitions, and context are ingested into the codex exec prompt argument as seen in SKILL.md.
Boundary markers: The skill documentation does not mention the use of delimiters or specific instructions to the sub-agent to ignore potential instructions embedded within the data.
Capability inventory: The skill uses the codex tool which, according to the instructions, can be granted workspace-write permissions, allowing it to modify files in the repository.
Sanitization: There are no documented steps for escaping or validating the content of the self-contained prompts before they are passed to the shell.

codex-delegate