The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection through the processing of untrusted repository data.
Ingestion points: Processes external git logs, file contents, and pull request metadata to build commit plans.
Boundary markers: Employs a mandatory commit-plan approval loop (Phase 4) and structured confirmation gates (Phase 0, Phase 8).
Capability inventory: Possesses significant capabilities including authenticated shell command execution (git, npm, cargo) and API interactions (gh, glab).
Sanitization: Validates generated commit messages against repository-defined hooks and provides a manual verification phase before the final push.
[COMMAND_EXECUTION]: The tool executes shell commands to perform git operations, interact with host APIs for branch and release management, and run repository-defined build or test scripts. These operations are transparently documented and subject to user approval.
[EXTERNAL_DOWNLOADS]: The skill references standard developer tooling from well-known registries (NPM, PyPI, Cargo) to support commit validation and release workflows, including tools like commitlint and semantic-release.

awesome-git-history-rebuild