The Agent Skills Directory

COMMAND_EXECUTION (SAFE): The skill uses local Git commands such as git status, git diff, and git commit to manage version control. These are standard operations for a developer tool and do not involve unauthorized shell access or administrative privileges.
DATA_EXFILTRATION (SAFE): No network operations, external URLs, or data transmission patterns were found. The skill does not communicate with external servers.
CREDENTIALS_UNSAFE (SAFE): No hardcoded API keys, tokens, or private credentials were found. The skill interacts with the local Git configuration which is expected behavior.
INDIRECT_PROMPT_INJECTION (LOW):
Ingestion points: Untrusted data enters the context via git diff analysis in the 'Analyze Changes' step.
Boundary markers: No specific delimiters are used to wrap the code diff output.
Capability inventory: The agent can execute git add and git commit (subprocess calls).
Sanitization: No explicit sanitization or filtering of the code content is performed. While an attacker could put malicious instructions in a code comment, the impact is limited to influencing the generated commit message text which the user is prompted to confirm.

git-commit