The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill processes untrusted data from git logs, diffs, and blame output which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: git log -p, git diff, and git blame output is read into the agent context.
Boundary markers: Absent; there are no specific instructions for the agent to ignore instructions embedded within the git data.
Capability inventory: The skill can perform git commit, git push, and git bisect run (script execution).
Sanitization: The skill proactively filters secret files (like .env) but does not sanitize the text content of files for prompt injection patterns.
[Command Execution] (LOW): The skill utilizes git bisect run ./test-script.sh, which executes a local script as part of the debugging process.
Evidence: Use of git bisect run in History Mode.
Context: This is a standard and powerful Git feature required for the skill's archaeology purpose. The severity is low because it targets local repository files rather than remote downloads, and is restricted by the intended primary purpose of the tool.

git-master