The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the GitHub CLI (gh) to interact with pull requests. It performs gh pr view to retrieve the current PR body, files, and commits, and gh pr edit to update the description. These operations are standard and aligned with the skill's documented purpose.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources (PR bodies and commit messages) when fetching PR details. This surface could allow for indirect prompt injection where instructions in the PR content influence the agent. The risk is mitigated by strict instructional formatting rules that prioritize user-facing behavior and plain-English descriptions.
Ingestion points: PR bodies, commit messages, and file diffs fetched via gh pr view as defined in SKILL.md.
Boundary markers: No explicit markers or instructions to ignore embedded commands are present.
Capability inventory: PR editing via gh pr edit, screenshot capturing via agent-browser, and file uploads via github-image-upload.
Sanitization: The skill does not perform validation or sanitization of the fetched PR content before processing it.

pr-descriptions