resolving-merge-conflicts
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read external data sources including commit messages, pull requests, and issue trackers to understand the context of conflicts. 1. Ingestion points: Git history, conflicting files, commit messages, pull requests, and issue/ticket descriptions in SKILL.md. 2. Boundary markers: No delimiters or ignore instructions specified. 3. Capability inventory: Execution of project-specific automated checks (tests, typechecking, formatting) via shell commands. 4. Sanitization: No explicit validation or filtering of content retrieved from external sources.
- [COMMAND_EXECUTION]: Step 4 directs the agent to discover and execute the project's automated checks (e.g., tests, typecheckers, and linters). This involves executing code defined within the project repository environment (SKILL.md).
Audit Metadata