The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the processing of git repository data.
Ingestion points: The agent reads output from git diff --cached and git log to generate commit messages and changelogs.
Boundary markers: There are no explicit markers or instructions defined to isolate the agent from instructions that might be embedded in code diffs or commit history.
Capability inventory: The agent can execute shell commands (npm, composer, gh), modify project files, and interact with the GitHub API to change repository settings.
Sanitization: No sanitization or filtering of the ingested git data is described in the skill instructions.
[COMMAND_EXECUTION]: The skill performs administrative repository management using the GitHub CLI.
It uses gh api to programmatically configure branch protection rules and update repository settings like squash merge defaults.
It automates the release process by creating tags and GitHub Releases using local metadata.
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of standard development dependencies from official registries.
It installs well-known tools for commit linting and git hooks, such as @commitlint/cli, husky, and captainhook/captainhook via NPM and Composer.
These downloads target established registries and represent standard industry practices for project configuration.

git-workflow