finishing-a-development-branch

Pass

Audited by Gen Agent Trust Hub on Mar 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes shell commands to run local test suites using npm, cargo, pytest, or go. It also performs standard branch management through git commands.- [EXTERNAL_DOWNLOADS]: The skill triggers network operations via git pull, git push, and the GitHub CLI (gh) to interact with remote repositories and manage Pull Requests.- [PROMPT_INJECTION]: There is an indirect injection surface where the agent ingests data from branch names, commit logs, and test results to populate a Pull Request template.
  • Ingestion points: Data is pulled from git branch metadata and command-line outputs of test tools.
  • Boundary markers: There are no explicit delimiters to isolate ingested data from agent instructions.
  • Capability inventory: The agent has the ability to execute shell commands and modify remote repository state.
  • Sanitization: No explicit sanitization or validation of the branch names or commit messages is performed before they are included in the PR body.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 8, 2026, 07:40 AM