The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): The skill is highly susceptible to indirect prompt injection because it processes the content of file diffs to determine its next actions.
Ingestion points: SKILL.md uses git diff --cached --stat and git diff --stat to populate context for the LLM.
Boundary markers: Absent. There are no delimiters or instructions to the model to ignore natural language instructions found within the code being analyzed.
Capability inventory: The skill has Bash(git :*) permissions and is instructed to perform git add ., git commit, and git push automatically.
Sanitization: Absent. The LLM processes raw diff statistics and potentially file contents without sanitization.
[Command Execution] (MEDIUM): The skill executes Git commands that modify and export data (pushing to a remote repository) without human oversight.
The 'NO INTERACTION' and 'AUTO-PUSH' rules in SKILL.md ensure that any malicious influence from an indirect prompt injection results in immediate, non-reversible actions in the repository.

commit