The Agent Skills Directory

[SAFE]: The skill implements a set of defensive instructions designed to ensure git operations are performed securely. It prioritizes user confirmation for sensitive actions and avoids patterns known to expose credentials.\n- [COMMAND_EXECUTION]: The skill uses the system git command to perform operations. The provided guardrails limit execution to safe, non-destructive patterns and require explicit confirmation for actions like pushing or deleting branches.\n- [DATA_EXFILTRATION]: The instructions explicitly prohibit the transmission of repository data to external services and warn against storing sensitive diffs (e.g., from .env files) in insecure local directories.\n- [PROMPT_INJECTION]: The 'Security Guardrails' section provides defensive instructions to the agent to prevent social engineering attacks that might attempt to bypass safety filters or ignore established security protocols.

git-helper-hardened