The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes project-specific files to determine execution logic.
Ingestion points: Reads CLAUDE.md and manifest files like package.json or requirements.txt from the local repository.
Boundary markers: Lacks explicit delimiters or instructions to ignore embedded commands within the ingested files.
Capability inventory: Employs multiple shell-based tools including git, npm, pip, cargo, go, and gh for environment setup and lifecycle management.
Sanitization: Content from configuration files is used to drive command execution paths without explicit sanitization.
[COMMAND_EXECUTION]: The skill automates development tasks using system-level CLI tools.
Evidence: Orchestrates git worktree for isolation, npm/pip/cargo/go for dependency management, and npm test/pytest for verification.
[EXTERNAL_DOWNLOADS]: The skill communicates with external git remotes and the GitHub platform.
Evidence: Synchronizes with remote repositories using git pull and git push, and interacts with GitHub APIs via the gh CLI tool.

git-workflow