git-undo-wizard
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by instructing the agent to perform Git operations based on user-provided parameters.
- Ingestion points: User input defining branch names, commit hashes, and file paths (SKILL.md Scenarios 1-10).
- Boundary markers: Absent; the instructions do not explicitly tell the agent to delimit or ignore instructions embedded in user-supplied strings.
- Capability inventory: The skill leverages the agent's ability to execute shell commands, specifically
gitbinaries (git reset,git revert,git checkout,git cherry-pick,git clean,git push). - Sanitization: Absent; there are no instructions to sanitize or validate the user-provided identifiers before interpolation into command strings.
- [COMMAND_EXECUTION]: The skill provides examples of destructive commands (
git reset --hard,git clean -fd) and history-rewriting commands (git push --force). While these are standard for Git recovery, they pose a risk of data loss if executed incorrectly or via malicious injection.
Audit Metadata