The Agent Skills Directory

[Prompt Injection] (LOW): The skill implements an LLM-based workflow to synthesize release notes from git commit history, creating an indirect prompt injection surface. -- Ingestion points: The synthesis process (referenced in scripts/synthesize-release-notes.mjs) consumes git commit messages which are untrusted external data. -- Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the provided configuration templates. -- Capability inventory: The setup includes GitHub Actions workflows with permissions to write contents, issues, and pull requests. -- Sanitization: The configuration does not specify sanitization or escaping of commit messages before they are processed by the LLM.
[External Downloads] (LOW): The skill installs multiple npm packages (e.g., semantic-release, commitlint, lefthook). While these are well-known development tools, they are downloaded from the public registry without pinned versions. The severity is LOW as these actions are central to the skill's primary purpose.
[Command Execution] (SAFE): The skill performs transparent setup tasks, including creating configuration files, installing git hooks, and managing GitHub secrets via the official CLI. These operations are appropriate for the tool's functionality.

changelog-setup