The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill retrieves data from GitHub pull request history and local templates to guide the generation of new PR content. This creates a surface for indirect prompt injection if existing PRs contain instructions intended to override agent behavior.
Ingestion points: scripts/ctx.sh and scripts/ctx.cmd fetch the title and body of recent pull requests using gh pr list and read .github/pull_request_template.md from the repository.
Boundary markers: The instructions lack explicit delimiters or warnings for the agent to treat ingested PR content as data rather than instructions.
Capability inventory: The skill utilizes gh pr create to make changes to the remote repository and xdg-open, open, or start to interact with the host operating system.
Sanitization: No sanitization is applied to the PR bodies or titles retrieved from the GitHub API before they are processed by the agent.
[COMMAND_EXECUTION]: The skill relies on several shell commands to function, including git for diffing and branch information, the GitHub CLI (gh) for label and PR management, and platform-specific commands like start, notepad, open, or xdg-open to present generated content to the user for review.

pr-objectively