The Agent Skills Directory

[COMMAND_EXECUTION]: The skill automatically stages all changes in the current directory using git add . via the Bash tool. While it performs subsequent safety checks on the staged files, the initial staging occurs without explicit user consent.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it analyzes content from the repository (via git diff) to generate commit messages. Malicious content within a file's diff could potentially influence the model's output or behavior.
Ingestion points: Code changes analyzed via git diff --cached in SKILL.md.
Boundary markers: Absent; there are no delimiters used to separate the untrusted diff content from the skill's instructions.
Capability inventory: Access to the Bash tool for local git operations and network pushes in SKILL.md.
Sanitization: Absent; the content of the diff is processed directly to generate a commit message summary.

git-cmsg