The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from the local repository environment to inform its workflow and generate outputs.
Ingestion points: Reads AGENTS.md, maintainer documentation, and the current branch's commit history and code diffs to generate PR summaries (SKILL.md).
Boundary markers: None explicitly defined for the file read operations.
Capability inventory: The skill has the capability to execute shell commands for testing, git operations, and GitHub CLI interactions.
Sanitization: PR bodies are constructed using heredocs (EOF), which mitigates basic command injection, but the LLM-generated content derived from the repository could still be manipulated by malicious content within the analyzed files.
[COMMAND_EXECUTION]: The skill relies on executing various system commands to fulfill its development workflow automation.
Tool usage: Uses git for branch switching, merging, and worktree management; gh for pull request creation; and language-specific test runners including npm, cargo, pytest, and go.
Execution pattern: Commands are largely static templates that incorporate variable branch names and programmatically generated PR descriptions.

finishing-a-development-branch