finish-work
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes multiple Git commands that can modify the local filesystem and remote repository state, including merging code, pushing to remote servers, and deleting branches/worktrees.
- Evidence:
SKILL.mdandCLEANUP.mdcontain instructions forgit merge,git push,git branch -D, andgit worktree remove --force. - [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external Git tools and processes outputs from other skills, which could potentially be manipulated to influence the agent's behavior.
- Ingestion points: Tool outputs from
git status --porcelain,git branch --show-current, and outcomes from thecode-reviewandverifying-completionskills. - Boundary markers: No explicit delimiters or boundary markers are defined for the interpolation of tool outputs or external skill data.
- Capability inventory: The skill possesses the capability to modify the repository via merges, push code to remote branches, and permanently delete branches and worktrees.
- Sanitization: No explicit sanitization or validation of Git branch names or status outputs is implemented before they are used in the construction of subsequent commands.
Audit Metadata