The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to interact with the local shell to perform GitHub repository operations. These include linking, syncing, and merging pull requests via the gh stack command set.
[PROMPT_INJECTION]: The skill processes repository data (branch names, PR metadata) which serves as an ingestion point for potential indirect prompt injection. Ingestion points: PR metadata and branch names fetched via gh pr view and the GitHub GraphQL API. Boundary markers: No explicit delimiter instructions are provided in the prompt. Capability inventory: The agent can execute potentially destructive shell commands like gh stack push or gh stack merge. Sanitization: The skill includes a specific check requiring the author.login to match exactly before automated linking occurs, which mitigates simple impersonation attempts.
[REMOTE_CODE_EXECUTION]: The skill relies on the gh CLI and the gh stack extension, both of which are official tools from a well-known and trusted vendor. It does not involve downloading or executing arbitrary scripts from third-party or untrusted servers.
[SAFE]: All operations are consistent with the skill's stated purpose of landing GitHub PR stacks. The logic favors official GitHub objects and APIs over manual branch manipulation, reducing the risk of state corruption.

merging-stacked-prs