The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local shell commands including gh, ls, head, and xxd. These tools are used for their intended functions: interacting with the GitHub CLI, navigating the filesystem, and generating unique temporary identifiers for issue body files.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data by listing existing issues and reading issue templates. This creates a potential surface for indirect prompt injection if the ingested data contains malicious instructions. However, the skill provides a security control by requiring the agent to display the draft to the user for manual review and confirmation before the issue is created.
Ingestion points: Output from gh issue list and content from .github/ISSUE_TEMPLATE/ files.
Boundary markers: The instructions lack specific text delimiters for untrusted data, but include a behavioral constraint for user review.
Capability inventory: The skill has the ability to write files to the .turbo/issue/ directory and create issues via the gh command-line tool.
Sanitization: The process relies on manual user verification of the drafted issue content to prevent automated submission of malicious payloads.

create-issue