The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes a command-line utility named codex-subagent. This tool appears to be the primary mechanism for the skill to delegate analysis tasks to a specialized agent.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data such as issue or pull request context, which could contain instructions intended to influence the subagent's behavior.
Ingestion points: The SKILL.md file defines a run command that accepts task context, scope, and issue/PR numbers provided by the user or external systems.
Boundary markers: The skill uses a quoted heredoc (<<'EOF') to wrap the input, which prevents the local shell from performing variable expansion on the content, providing a basic structural boundary.
Capability inventory: The skill is capable of executing the codex-subagent CLI tool with the provided context.
Sanitization: No explicit sanitization, filtering, or instruction-override protection is implemented within the skill for the external context it ingests.

comment-analyzer