The Agent Skills Directory

[SAFE]: The skill is designed to streamline Git operations and GitHub pull request management. It relies on standard tools (git and gh CLI) and does not contain any obfuscated code, credential harvesting, or unauthorized network operations.
[COMMAND_EXECUTION]: Executes shell commands via the Bash tool to perform repository actions such as staging changes, committing with conventional formats, merging branches, and interacting with GitHub PRs. All executed commands are transparently defined within the skill's scenarios.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data including commit histories (via git log), pull request templates, and existing PR descriptions to assist the user in writing summaries. While this represents a potential surface where an attacker could influence the agent's writing by placing instructions in commit messages, the impact is limited to the generated text and does not affect the safety of the execution environment.
Ingestion points: SKILL.md (processes output from git log, .github/pull_request_template.md, and gh pr view)
Boundary markers: None explicitly defined in the bash snippets.
Capability inventory: Execution of git and gh commands via Bash.
Sanitization: Relies on the agent's internal handling of command arguments.

git-flow