git-audit
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local Git commands using a Python script (
scripts/inspect_repository.py). It follows security best practices by passing arguments as a list tosubprocess.Popen(preventing shell injection) and stripping sensitive environment variables likeGIT_DIRandGIT_CONFIG_PARAMETERSto prevent unauthorized configuration. - [INDIRECT_PROMPT_INJECTION]: The skill reads branch names and metadata files (e.g.,
.gitflow/state/activation.json) from untrusted repositories. While this is an ingestion point for external data, the skill's read-only nature and structured reporting minimize the risk of malicious instructions affecting the agent. - Ingestion points: Git branch names, commit history, and project configuration files.
- Boundary markers: Results are reported in structured JSON format.
- Capability inventory: Limited to read-only Git status commands.
- Sanitization: Python script handles output as strings and limits total output size to 4MB to prevent resource exhaustion.
Audit Metadata