The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a series of shell commands using git and gh (GitHub CLI) to manage local and remote branches. These commands include listing branches, checking commit counts, pushing branches to origin, and creating/commenting on Pull Requests.
[DATA_EXFILTRATION]: The skill performs network operations specifically targeting GitHub through the git push and gh pr commands. This is consistent with the skill's primary purpose of PR automation on a well-known service. No sensitive local configuration files (e.g., .ssh/id_rsa, .env) are accessed or transmitted.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from the local environment (specifically git branch names and PR numbers) and interpolates them into PR titles, bodies, and comments. This represents a minor indirect injection surface where malicious branch names could influence the agent's output, though the risk is minimized by the intended use in a controlled developer environment.
[REMOTE_CODE_EXECUTION]: The skill references external dependencies (e.g., sanctum:pr-prep) which are part of the same ecosystem/vendor. It does not download or execute arbitrary scripts from untrusted remote sources.

stack-push