The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands and the GitHub CLI (gh) to perform repository operations, including branch management, stashing, and PR creation. These actions are fundamental to its intended purpose.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the local repository and GitHub environment (such as branch names, commit messages, and issue links) to generate a plan. While this is a potential attack surface, it is mitigated by a mandatory human approval step before any commands are executed. Ingestion points: Git branch names, Linear/GitHub issue links, and file diffs. Boundary markers: The skill enforces a 'Wait for explicit approval' rule. Capability inventory: Access to git and gh CLI commands. Sanitization: Not specified.\n- [SAFE]: The skill demonstrates safe practices by creating recoverable backups (git stash and custom backup references) before modifications and explicitly forbidding destructive git operations without user consent.

to-pr-stack