The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to execute various shell commands for version control and project management, including git (for commits, branches, and merging), npm/npx (for package installation and running automation tools), and the GitHub CLI gh (for creating and merging pull requests).
[DYNAMIC_EXECUTION]: As part of its configuration guidance, the skill generates shell scripts for git hooks, specifically writing commands to .husky/pre-commit and .husky/commit-msg to automate linting and commit message validation during the local development lifecycle.
[INDIRECT_PROMPT_INJECTION]: The skill defines a surface for processing external data by interpolating user-provided inputs like ticket IDs and commit descriptions into PR and commit templates.
Ingestion points: Data enters the system via commit subjects and PR body content placeholders defined in SKILL.md and reference/commit-examples.md.
Boundary markers: Instructions are structured using standard Markdown headers and XML-like tags such as <conventional_commits> and <pr_templates>.
Capability inventory: The skill performs file writes to the project's .husky/ directory and agent-specific telemetry paths (~/.claude/skill-analytics/), and executes system commands through git, npm, and gh.
Sanitization: There are no explicit instructions for escaping or validating user-supplied content before it is processed into final commit or PR messages.

git-workflow