The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted external content (GitHub issues, discussions, reports) to generate pull request metadata, creating a surface for indirect prompt injection. * Ingestion points: The skill explicitly uses user-supplied reports, issues, and discussions as primary context sources in SKILL.md. * Boundary markers: There are no instructions to use delimiters or ignore embedded instructions within the ingested external data. * Capability inventory: The skill executes git and gh (GitHub CLI) commands to manage branches, commits, and pull requests. * Sanitization: The use of --body-file for gh pr create prevents shell-level injection in the command itself, though the agent remains susceptible to instructions within the data content.
[COMMAND_EXECUTION]: The skill utilizes git and gh commands to perform its core functions. These are standard tools for the intended environment and are used with appropriate flags (e.g., --body-file) to mitigate common command-line risks. Additionally, the skill instructs the agent to update its own label list within SKILL.md if existing labels are found to be invalid, which is a form of self-modification.

create-pr