commit
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses Git commands (
git status,git diff,git add,git commit) to manage the repository state. It correctly utilizes HEREDOC syntax (cat <<'EOF') for commit messages, which prevents the shell from interpreting special characters or executing commands embedded in the commit text. - [INDIRECT_PROMPT_INJECTION]: The skill analyzes external data in the form of code diffs and file contents. It follows a mandatory evidence chain to mitigate risks:
- Ingestion points: Reads
git statusandgit diffoutput to understand changes. - Boundary markers: Distinguishes between read-only 'message-only' mode and 'execute-commit' mode to prevent unintended state changes.
- Capability inventory: Has the ability to stage files (
git add), commit changes (git commit), and modify the.gitignorefile. - Sanitization: Implements strict checks for sensitive file patterns (e.g.,
.env, keys, tokens) and requires explicit user confirmation if such files are detected before staging.
- Ingestion points: Reads
- [DATA_EXPOSURE]: The skill explicitly identifies and excludes sensitive file paths (credentials, tokens, private keys) from the commit process. It implements logic to automatically append these patterns to
.gitignoreif they are missing, preventing future accidental exposure.
Audit Metadata