cleaning-up-branches
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes git commands using the Bash tool to manage local and remote branches. It implements safety features such as dry-run flags and mandatory user confirmation before destructive operations like branch deletion.
- [INDIRECT_PROMPT_INJECTION]: The skill reads branch names and commit metadata from the repository, which constitutes an attack surface for indirect prompt injection. 1. Ingestion points: The skill displays outputs from git branch, git branch -r, and git for-each-ref (SKILL.md, Steps 4-8). 2. Boundary markers: Sections are delimited by headers, but individual branch names and metadata strings are not wrapped in safety delimiters. 3. Capability inventory: The skill has permissions to delete local branches (git branch -d/-D) and remote branches (git push origin --delete). 4. Sanitization: There is no evidence that branch names or commit messages are sanitized before being interpolated into the agent's context.
Audit Metadata