The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from repository history (git log) and code differences (git diff) to generate automated commit messages and PR summaries. This creates a surface where malicious content in a repository's history could influence the agent's behavior.
Ingestion points: The files pr-analyze.md and commit.md execute commands to read git logs and diffs, passing the output into the agent's context for summarization.
Boundary markers: Absent. The instructions do not define clear delimiters or use specialized prompts to isolate the external git data from the agent's core instructions.
Capability inventory: The skill possesses capabilities to modify the repository and interact with remote services, including git commit, git push (in push.md and pr.md), and gh pr create (in pr.md).
Sanitization: Absent. There is no evidence of filtering or sanitizing the content of commit messages or diffs before they are processed by the AI.
[COMMAND_EXECUTION]: The skill is designed to execute a wide variety of shell commands using the git and gh binaries. These operations include branch management, history rewriting via git reset --soft and git rebase, and authenticated network operations via the GitHub CLI. While these are necessary for the skill's functionality, they provide the mechanism through which an injection attack could potentially reach the underlying system or remote repository.

custom-skills-git-commit