rebase
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides a set of local Bash scripts that wrap standard Git commands, ensuring rebase operations are performed under controlled, pinned settings to avoid unpredictable behavior from operator preferences.- [SAFE]: A mandatory guard hook ('guard-rebase.sh') is implemented to block potentially hazardous commands such as bare 'git push --force', 'git stash pop', and manual rebase continuation, forcing the agent to use safer alternatives like '--force-with-lease'.- [SAFE]: The skill uses dynamic context injection in the SKILL.md file to safely gather repository status (via 'preflight.sh') at load time, which is a legitimate use of the platform's features for situational awareness.- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Git history (logs, diffs) and file contents (conflict markers). While this is an attack surface, the risk is minimized by the skill's reliance on structured scripts and its primary purpose as a Git management tool. \n
- Ingestion points: 'scripts/preflight.sh' (log analysis), 'scripts/rebase-status.sh' (status and diff analysis), and 'scripts/marker-scan.sh' (scanning files for conflict markers).\n
- Boundary markers: The skill provides a specific workflow and scripts to interpret Git output, reducing the likelihood of the agent following instructions embedded in commit messages or code.\n
- Capability inventory: The skill can execute Git commands, Bash scripts, and Mise tasks.\n
- Sanitization: The 'guard-rebase.sh' script serves as an automated sanitizer, enforcing safe command usage for all Git-related Bash operations.
Audit Metadata