branch-cleaner
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes Git and GitHub CLI commands to manage repository branches.\n
- Evidence: Use of commands such as
git branch,git push --delete,git status, andghinSKILL.mdto audit and prune branches.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from local Git metadata and branch names, which could be manipulated to influence the agent.\n - Ingestion points: Command output from
git status,git branch, andgit for-each-refis processed by the agent inSKILL.md.\n - Boundary markers: No explicit prompt delimiters are defined to separate tool output from instructions, although specific Git flags like
--porcelainare used for structured output.\n - Capability inventory: The skill includes destructive capabilities such as local and remote branch deletion (
git branch -d,git push --delete).\n - Sanitization: The instructions include strict safety protocols, including a prohibition on deleting via unresolved globs and a requirement for explicit user approval and state re-verification immediately before deletion.
Audit Metadata