git-workflow

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes various git commands and local Python scripts. It uses a secure heredoc pattern (cat <<'EOF') when executing git commit to prevent shell injection vulnerabilities from the generated commit message text.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it processes untrusted data from the repository (git logs, diffs, and source code) to generate summaries and recommendations.
  • Ingestion points: Data enters the agent context through git diff --cached, git log, and by reading conflicted or archaeological source files.
  • Boundary markers: The skill uses structured JSON output from its helper scripts and shell heredocs to isolate data from commands.
  • Capability inventory: The skill can perform git commit, git bisect (which checks out different commits), and read any file within the repository.
  • Sanitization: Input from git commands is parsed into structured JSON by commit-parser.py and diff-summarizer.py before being processed by the LLM.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 02:19 AM