The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands to interact with the repository and post data to GitHub.
Evidence: It calls git ls-files to understand project structure and gh issue create to post the final issue.
Mitigation: The skill instructions include a mandatory "best practice" to never post without user confirmation, requiring a full preview of the title and body before execution.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface where instructions embedded in codebase files or user descriptions could influence the agent's behavior.
Ingestion points: Reads local files (README, entry-point files) and user-supplied descriptions.
Boundary markers: None explicitly defined in the prompt template, but the skill enforces a human review stage.
Capability inventory: Can read arbitrary files in the repository and execute gh issue create with generated arguments.
Sanitization: Relies on the user to review the preview [yes / edit / cancel] before the command is executed.

create-issue