git-merge

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes standard, non-privileged git commands (git status, git diff, git log) to identify merge conflicts and retrieve version history. This behavior is consistent with the stated purpose of resolving source control conflicts.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from incoming git branches and project design artifacts to resolve conflicts.
  • Ingestion points: Conflicted files (aidlc-state.md, audit.md) and application code files as described in SKILL.md and the referenced merge strategy documents.
  • Boundary markers: The instructions do not specify explicit delimiters or "ignore previous instructions" warnings when the agent reads external file content for analysis.
  • Capability inventory: The skill has access to shell commands (git) and file system read/write operations within the project directory.
  • Sanitization: There is no explicit sanitization of the content found within the conflicted files before the agent parses them for resolution logic.
  • Note: This represents a documented attack surface inherent to tools that analyze and merge external source code; however, no malicious intent was observed.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 09:22 AM
Security Audit — agent-trust-hub — git-merge