監査履歴

Static external-command findings are confirmed as intentional Git usage with fixed argument lists, not shell injection. The weak-cryptography findings are false positives on prose and AI instruction text. A medium risk remains because full staged diff content is placed into the model context, where malicious diff text could influence the drafted message.

This skill is a legitimate git utility. Static scanner found 20 patterns but ALL are false positives. The scanner misidentified Markdown backticks as shell execution, JSON text as cryptographic algorithms, and a source URL as a network vulnerability. The code only runs hardcoded read-only git commands (status, diff, rev-parse) with argument lists and timeouts.

This skill is a legitimate git utility. Static scanner found 20 patterns but ALL are false positives. The scanner misidentified Markdown backticks as shell execution, JSON text as cryptographic algorithms, and a source URL as a network vulnerability. The code only runs hardcoded read-only git commands (status, diff, rev-parse) with argument lists and timeouts.

This skill is a legitimate git utility that analyzes staged changes to help draft commit messages. It uses subprocess to run read-only git commands (status, diff, rev-parse) with hardcoded arguments. No network calls, no arbitrary command execution, no file access outside the git repository.

This skill is a legitimate git utility that analyzes staged changes to help draft commit messages. It uses subprocess to run read-only git commands (status, diff, rev-parse) with hardcoded arguments. No network calls, no arbitrary command execution, no file access outside the git repository.

This skill is a legitimate git utility that analyzes staged changes to help draft commit messages. It uses subprocess to run read-only git commands (status, diff, rev-parse) with hardcoded arguments. No network calls, no arbitrary command execution, no file access outside the git repository.