The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple shell commands to interact with the local git repository and run project-specific test suites.
Evidence: Instructions include running npm test, cargo test, pytest, and go test for verification, as well as git merge, git push, and git worktree remove for integration and cleanup in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill's behavior is guided by data ingested from the development environment and external tool outputs, which constitutes a potential attack surface.
Ingestion points: Git metadata (branch names, worktree paths) retrieved via git rev-parse and output from project test suites as described in SKILL.md.
Boundary markers: The skill implements a strict decision menu and requires explicit user confirmation before proceeding with integration or deletion.
Capability inventory: The skill can modify the local file system (merging, branch deletion), perform network operations (pushing to remote), and execute project-defined test code as detailed in SKILL.md.
Sanitization: The skill relies on explicit user confirmation for the base branch identity and requires a specific keyword ("discard") before performing destructive workspace deletions.

finishing-a-development-branch