The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a wide variety of shell commands, including Git operations (git worktree add, git worktree list, git status), file operations (cp .env.example .env, git worktree remove), and ecosystem-specific setup/test commands.
[INDIRECT_PROMPT_INJECTION]: The skill evaluates and reacts to data found within the local project environment to determine its workflow.
Ingestion points: The skill reads git worktree list output, checks CLAUDE.md for specific configuration keys, and scans the directory structure for lockfiles and configuration files (e.g., package.json, pyproject.toml, go.mod).
Boundary markers: No specific boundary markers or warnings regarding embedded instructions in these files are present.
Capability inventory: The skill possesses the capability to execute arbitrary shell commands via multiple package managers (npm, pnpm, yarn, pip, poetry, go, cargo, bundle, composer) and test runners (pytest, npm test, etc.) as documented in the Setup Detection and Test Verification phases.
Sanitization: No explicit sanitization or validation of the content found in the external configuration files is performed before it influences command selection.
[DYNAMIC_EXECUTION]: The skill employs a dynamic decision logic (Phase 4: Setup Detection) where the agent identifies the project's ecosystem and automatically executes the associated installation commands. This runtime selection of executable paths based on detected project files is a form of dynamic instruction generation.

using-git-worktrees