resolving-merge-conflicts

Pass

Audited by Gen Agent Trust Hub on Aug 24, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill directs the agent to ingest and interpret external data that could contain malicious instructions designed to influence the agent's behavior during conflict resolution.
  • Ingestion points: The agent is instructed in steps 2 and 3 to read Git commit messages, code history, Pull Request (PR) descriptions, Issues, task cards, and requirement documents.
  • Boundary markers: There are no instructions to wrap these external inputs in delimiters or to ignore potential instructions embedded within the metadata or descriptions of these resources.
  • Capability inventory: The agent has the capability to write to the file system (resolving conflicts) and execute arbitrary project commands (running tests, type checks, and formatters) as described in step 4.
  • Sanitization: No specific sanitization, filtering, or validation of the content retrieved from PRs or Issues is mentioned before the agent uses that information to make decisions on code changes.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 24, 2026, 03:54 PM
Security Audit — agent-trust-hub — resolving-merge-conflicts