skills/basher83/lunar-claude/git-ops/Gen Agent Trust Hub

git-ops

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill translates natural language into git operations and transmits user phrases to a feedback tool for vocabulary learning. 1. Ingestion points: User requests such as 'back it up' or 'save my work' in SKILL.md. 2. Boundary markers: None defined; the agent relies on an intent mapping table. 3. Capability inventory: Invokes blake-os:git-commit, blake-os:git-sync, and blake-os:feedback, and executes shell-based git commands. 4. Sanitization: Proactively forbids staging sensitive files like .env or keys and prohibits commands like force-push, reset, and rebase.
  • [COMMAND_EXECUTION]: Direct execution of git commands to retrieve repository status. Evidence: hardcoded strings for git fetch origin, git status, and git rev-list in SKILL.md. These commands are informational and do not incorporate direct user parameters into the shell execution string.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 04:54 PM
Security Audit — agent-trust-hub — git-ops