finishing-a-development-branch
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes standard system commands for development workflows, including testing runners (npm, cargo, pytest, go), git management (merge-base, checkout, pull, merge, branch, push, worktree), and GitHub CLI (gh pr create).
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests and processes untrusted data from the working environment.
- Ingestion points: Instructions in SKILL.md direct the agent to read branch names, commit lists, and documentation files (CLAUDE.md).
- Boundary markers: None identified; untrusted strings are interpolated directly into command arguments and subagent prompts.
- Capability inventory: SKILL.md defines steps that execute shell commands and invoke external subagents.
- Sanitization: No sanitization or escaping mechanisms are implemented for external data before processing.
- [SAFE]: The skill interacts with the ed3d-extending-claude:project-claude-librarian subagent to update project documentation, which is a resource belonging to the skill's author.
Audit Metadata