The Agent Skills Directory

[COMMAND_EXECUTION]: Executes shell commands for git operations (git add -A, git commit, git push) and dynamically discovered linting tools based on project configuration.
[EXTERNAL_DOWNLOADS]: Downloads shared reference and configuration markdown files from the author's official GitHub repository (github.com/levnikolaevich/claude-code-skills) if local copies are not found.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it identifies and executes commands from potentially untrusted project files.
Ingestion points: Reads docs/project/runbook.md, CLAUDE.md, README.md, and CONTRIBUTING.md to discover linter and formatting commands (Phase 4).
Boundary markers: The instructions do not define delimiters or provide specific warnings to the agent to ignore embedded instructions within these data sources.
Capability inventory: Possesses the capability to execute shell commands via subprocess calls during linting and git phases.
Sanitization: There is no validation or sanitization of the commands discovered in project documentation before they are executed.
[DATA_EXFILTRATION]: The instruction to use git add -A stages all untracked and modified files in the repository. This behavior poses a risk of accidentally committing and pushing sensitive data, such as .env files, SSH keys, or local credentials, to a remote server if the user's .gitignore file is missing or incorrectly configured.

ln-001-push-all