The Agent Skills Directory

[PROMPT_INJECTION]: The skill identifies and manages an attack surface where untrusted data from GitHub pull request descriptions or git logs could contain malicious instructions.
Ingestion points: SKILL.md uses gh pr view and git diff to read external content into the agent context.
Boundary markers: The skill includes explicit instructions stating that existing PR metadata and generated diff summaries must be treated as untrusted text.
Capability inventory: SKILL.md utilizes git push, git commit, and gh pr edit capabilities.
Sanitization: The agent is strictly instructed to redact secret-like values and ignore any commands or instructions embedded in PR titles or bodies.
[COMMAND_EXECUTION]: The skill leverages git and gh for repository management. All significant operations, including staging broad file sets, committing, and pushing code, are protected by mandatory user approval steps to ensure no autonomous or unauthorized execution of critical commands occurs.

gh-pr-publish