NYC

recovering-from-bad-git-state

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • COMMAND_EXECUTION (LOW): The skill documentation includes commands for manual repository repair, such as rm -rf .git/worktrees/<worktree-name> and direct modification of .git/HEAD. These operations bypass Git's internal safety checks and could lead to unintended data deletion if provided with incorrect parameters.
  • PROMPT_INJECTION (LOW): The skill exhibits an indirect prompt injection surface by ingesting data from external tools like gh pr view and Git status commands and interpolating them into destructive actions.
  • Ingestion points: git worktree list, gh pr view output (SKILL.md).
  • Boundary markers: Absent.
  • Capability inventory: rm -rf, git branch -D, git push --delete (SKILL.md).
  • Sanitization: Absent.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:24 PM