The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill is designed to ingest and process untrusted data from git logs, PR titles, and issue descriptions which could contain malicious instructions.
Ingestion points: git log output, PR titles, and issue references listed in the 'Commit Analysis' section of SKILL.md.
Boundary markers: Absent. There are no specific delimiters or instructions for the agent to ignore prompt-like content within commit messages.
Capability inventory: Shell command execution via git, npm, cargo, and gh CLI tools as documented in SKILL.md.
Sanitization: Absent. The skill provides no logic for escaping or validating the content of commit messages before processing.
[Command Execution] (SAFE): The skill uses standard git history and GitHub CLI commands (git log, gh release create) which are necessary for its primary functionality.
[External Downloads] (SAFE): Mentions installation of conventional-changelog-cli and git-cliff from official registries (npm and crates.io). These are established development tools relevant to the skill's purpose.

changelog-writer