The Agent Skills Directory

[SAFE]: The skill is purely instructional, providing best-practice guidance for Git workflows such as trunk-based development and release branching. It prioritizes repository-local rules and industry standards (Semantic Versioning, Conventional Commits).
[SAFE]: Security-conscious design is evident in the instructions, which mandate the use of read-only Git commands for initial inspection and explicitly state that destructive or external mutations (like force-pushing or creating tags) require explicit user consent.
[INDIRECT_PROMPT_INJECTION]: While the skill processes external data (repository contribution documents and Git metadata), this is inherent to its function. The risk is minimized by instructions to follow established local constraints and the requirement for user approval for state-changing actions.
Ingestion points: Local repository configuration and documentation (SKILL.md).
Boundary markers: None explicitly defined.
Capability inventory: Git command execution and workflow documentation (SKILL.md).
Sanitization: None explicitly defined, but behavior is constrained by user intent requirements.

git-workflow