repo-audit

Pass

Audited by Gen Agent Trust Hub on Nov 22, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes Git commit history (messages, diffs, and author metadata) which are untrusted external inputs. If a repository being audited contains specially crafted commit messages or content designed to influence an AI, it could lead to the agent deviating from its intended behavior.
  • Ingestion points: Git log output retrieved in scripts/hotfiles.sh, scripts/ownership.sh, and scripts/secret-scan.sh is returned to the agent's context for analysis and reporting.
  • Boundary markers: The output is generated as plain text tables, CSV, or JSON. There are no explicit boundary markers or system instructions provided to the agent to treat the analyzed Git metadata as untrusted data or to ignore embedded instructions.
  • Capability inventory: While the scripts themselves are read-only and lack network or file-writing capabilities, the agent interpreting the results may have access to other powerful tools (e.g., file system modification, shell execution) that an injection could attempt to exploit.
  • Sanitization: The scripts/secret-scan.sh script performs masking (redaction) of detected secrets to prevent credential exposure in the logs, but it does not filter or sanitize commit messages or other textual fields for prompt injection patterns.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 22, 2026, 01:12 AM