The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various git and GitHub CLI commands to manage the release workflow. This includes commands to read logs (
git log ${LAST_TAG}..HEAD --oneline), check diffs (
git diff ${LAST_TAG}..HEAD --stat), and perform write operations like tagging (
git tag v{VERSION}), committing (
git commit -m "..."), and pushing (
git push origin ...) to remote repositories. While intended for release automation, these commands provide the agent with high-level access to the repository's history and configuration.
[INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted data from git history and GitHub PRs without explicit sanitization or boundary markers.
Ingestion points: The skill ingests external content from git commit messages via git log (Step 2) and PR metadata via gh pr view (Step 4).
Boundary markers: There are no specific instructions or delimiters mentioned to separate untrusted commit/PR content from the agent's instructions, increasing the risk of the LLM following malicious commands embedded in history.
Capability inventory: The skill possesses powerful capabilities including the ability to modify project files (Step 7), create new commits/tags (Step 9), and push changes to remote servers (Step 9).
Sanitization: The instructions do not specify any validation or filtering for commit messages or PR descriptions before they are used to generate changelog entries or determine version bumps.

release-skills